id
large_stringlengths
9
16
submitter
large_stringlengths
3
64
authors
large_stringlengths
5
20.2k
title
large_stringlengths
5
245
comments
large_stringlengths
1
482
journal-ref
large_stringlengths
4
382
doi
large_stringlengths
9
151
report-no
large_stringlengths
2
509
categories
large_stringlengths
5
78
license
large_stringclasses
9 values
abstract
large_stringlengths
83
3.55k
versions
large listlengths
1
21
update_date
timestamp[ms]date
2007-05-23 00:00:00
2025-05-30 00:00:00
authors_parsed
large listlengths
1
1.35k
prompt
large_stringlengths
166
3.64k
label
stringclasses
2 values
prob
float64
0.5
0.73
1206.1270
Benjamin Recht
Victor Bittorf and Benjamin Recht and Christopher Re and Joel A. Tropp
Factoring nonnegative matrices with linear programs
17 pages, 10 figures. Modified theorem statement for robust recovery conditions. Revised proof techniques to make arguments more elementary. Results on robustness when rows are duplicated have been superseded by arxiv.org/1211.6687
null
null
null
math.OC cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper describes a new approach, based on linear programming, for computing nonnegative matrix factorizations (NMFs). The key idea is a data-driven model for the factorization where the most salient features in the data are used to express the remaining features. More precisely, given a data matrix X, the algorithm identifies a matrix C such that X approximately equals CX and some linear constraints. The constraints are chosen to ensure that the matrix C selects features; these features can then be used to find a low-rank NMF of X. A theoretical analysis demonstrates that this approach has guarantees similar to those of the recent NMF algorithm of Arora et al. (2012). In contrast with this earlier work, the proposed method extends to more general noise models and leads to efficient, scalable algorithms. Experiments with synthetic and real datasets provide evidence that the new approach is also superior in practice. An optimized C++ implementation can factor a multigigabyte matrix in a matter of minutes.
[ { "version": "v1", "created": "Wed, 6 Jun 2012 16:42:27 GMT" }, { "version": "v2", "created": "Sat, 2 Feb 2013 23:40:56 GMT" } ]
2013-02-05T00:00:00
[ [ "Bittorf", "Victor", "" ], [ "Recht", "Benjamin", "" ], [ "Re", "Christopher", "" ], [ "Tropp", "Joel A.", "" ] ]
TITLE: Factoring nonnegative matrices with linear programs ABSTRACT: This paper describes a new approach, based on linear programming, for computing nonnegative matrix factorizations (NMFs). The key idea is a data-driven model for the factorization where the most salient features in the data are used to express the remaining features. More precisely, given a data matrix X, the algorithm identifies a matrix C such that X approximately equals CX and some linear constraints. The constraints are chosen to ensure that the matrix C selects features; these features can then be used to find a low-rank NMF of X. A theoretical analysis demonstrates that this approach has guarantees similar to those of the recent NMF algorithm of Arora et al. (2012). In contrast with this earlier work, the proposed method extends to more general noise models and leads to efficient, scalable algorithms. Experiments with synthetic and real datasets provide evidence that the new approach is also superior in practice. An optimized C++ implementation can factor a multigigabyte matrix in a matter of minutes.
no_new_dataset
0.708023
1302.0413
Catarina Moreira
Catarina Moreira and P\'avel Calado and Bruno Martins
Learning to Rank for Expert Search in Digital Libraries of Academic Publications
null
Progress in Artificial Intelligence, Lecture Notes in Computer Science, Springer Berlin Heidelberg. In Proceedings of the 15th Portuguese Conference on Artificial Intelligence, 2011
null
null
cs.IR cs.DL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The task of expert finding has been getting increasing attention in information retrieval literature. However, the current state-of-the-art is still lacking in principled approaches for combining different sources of evidence in an optimal way. This paper explores the usage of learning to rank methods as a principled approach for combining multiple estimators of expertise, derived from the textual contents, from the graph-structure with the citation patterns for the community of experts, and from profile information about the experts. Experiments made over a dataset of academic publications, for the area of Computer Science, attest for the adequacy of the proposed approaches.
[ { "version": "v1", "created": "Sat, 2 Feb 2013 18:36:08 GMT" } ]
2013-02-05T00:00:00
[ [ "Moreira", "Catarina", "" ], [ "Calado", "Pável", "" ], [ "Martins", "Bruno", "" ] ]
TITLE: Learning to Rank for Expert Search in Digital Libraries of Academic Publications ABSTRACT: The task of expert finding has been getting increasing attention in information retrieval literature. However, the current state-of-the-art is still lacking in principled approaches for combining different sources of evidence in an optimal way. This paper explores the usage of learning to rank methods as a principled approach for combining multiple estimators of expertise, derived from the textual contents, from the graph-structure with the citation patterns for the community of experts, and from profile information about the experts. Experiments made over a dataset of academic publications, for the area of Computer Science, attest for the adequacy of the proposed approaches.
no_new_dataset
0.709265
1302.0540
Harris Georgiou
Harris V. Georgiou, Michael E. Mavroforakis
A game-theoretic framework for classifier ensembles using weighted majority voting with local accuracy estimates
21 pages, 9 tables, 1 figure, 68 references
null
null
null
cs.LG
http://creativecommons.org/licenses/by-nc-sa/3.0/
In this paper, a novel approach for the optimal combination of binary classifiers is proposed. The classifier combination problem is approached from a Game Theory perspective. The proposed framework of adapted weighted majority rules (WMR) is tested against common rank-based, Bayesian and simple majority models, as well as two soft-output averaging rules. Experiments with ensembles of Support Vector Machines (SVM), Ordinary Binary Tree Classifiers (OBTC) and weighted k-nearest-neighbor (w/k-NN) models on benchmark datasets indicate that this new adaptive WMR model, employing local accuracy estimators and the analytically computed optimal weights outperform all the other simple combination rules.
[ { "version": "v1", "created": "Sun, 3 Feb 2013 22:12:52 GMT" } ]
2013-02-05T00:00:00
[ [ "Georgiou", "Harris V.", "" ], [ "Mavroforakis", "Michael E.", "" ] ]
TITLE: A game-theoretic framework for classifier ensembles using weighted majority voting with local accuracy estimates ABSTRACT: In this paper, a novel approach for the optimal combination of binary classifiers is proposed. The classifier combination problem is approached from a Game Theory perspective. The proposed framework of adapted weighted majority rules (WMR) is tested against common rank-based, Bayesian and simple majority models, as well as two soft-output averaging rules. Experiments with ensembles of Support Vector Machines (SVM), Ordinary Binary Tree Classifiers (OBTC) and weighted k-nearest-neighbor (w/k-NN) models on benchmark datasets indicate that this new adaptive WMR model, employing local accuracy estimators and the analytically computed optimal weights outperform all the other simple combination rules.
no_new_dataset
0.709837
1302.0739
Conrad Lee
Conrad Lee, P\'adraig Cunningham
Benchmarking community detection methods on social media data
null
null
null
null
cs.SI physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Benchmarking the performance of community detection methods on empirical social network data has been identified as critical for improving these methods. In particular, while most current research focuses on detecting communities in data that has been digitally extracted from large social media and telecommunications services, most evaluation of this research is based on small, hand-curated datasets. We argue that these two types of networks differ so significantly that by evaluating algorithms solely on the former, we know little about how well they perform on the latter. To address this problem, we consider the difficulties that arise in constructing benchmarks based on digitally extracted network data, and propose a task-based strategy which we feel addresses these difficulties. To demonstrate that our scheme is effective, we use it to carry out a substantial benchmark based on Facebook data. The benchmark reveals that some of the most popular algorithms fail to detect fine-grained community structure.
[ { "version": "v1", "created": "Mon, 4 Feb 2013 16:12:22 GMT" } ]
2013-02-05T00:00:00
[ [ "Lee", "Conrad", "" ], [ "Cunningham", "Pádraig", "" ] ]
TITLE: Benchmarking community detection methods on social media data ABSTRACT: Benchmarking the performance of community detection methods on empirical social network data has been identified as critical for improving these methods. In particular, while most current research focuses on detecting communities in data that has been digitally extracted from large social media and telecommunications services, most evaluation of this research is based on small, hand-curated datasets. We argue that these two types of networks differ so significantly that by evaluating algorithms solely on the former, we know little about how well they perform on the latter. To address this problem, we consider the difficulties that arise in constructing benchmarks based on digitally extracted network data, and propose a task-based strategy which we feel addresses these difficulties. To demonstrate that our scheme is effective, we use it to carry out a substantial benchmark based on Facebook data. The benchmark reveals that some of the most popular algorithms fail to detect fine-grained community structure.
no_new_dataset
0.632077
1110.0305
Atieh Mirshahvalad
Atieh Mirshahvalad, Johan Lindholm, Mattias Derlen, Martin Rosvall
Significant communities in large sparse networks
7 pages, 7 figures
null
10.1371/journal.pone.0033721
null
physics.soc-ph cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Researchers use community-detection algorithms to reveal large-scale organization in biological and social networks, but community detection is useful only if the communities are significant and not a result of noisy data. To assess the statistical significance of the network communities, or the robustness of the detected structure, one approach is to perturb the network structure by removing links and measure how much the communities change. However, perturbing sparse networks is challenging because they are inherently sensitive; they shatter easily if links are removed. Here we propose a simple method to perturb sparse networks and assess the significance of their communities. We generate resampled networks by adding extra links based on local information, then we aggregate the information from multiple resampled networks to find a coarse-grained description of significant clusters. In addition to testing our method on benchmark networks, we use our method on the sparse network of the European Court of Justice (ECJ) case law, to detect significant and insignificant areas of law. We use our significance analysis to draw a map of the ECJ case law network that reveals the relations between the areas of law.
[ { "version": "v1", "created": "Mon, 3 Oct 2011 09:07:22 GMT" }, { "version": "v2", "created": "Fri, 1 Feb 2013 10:23:28 GMT" } ]
2013-02-04T00:00:00
[ [ "Mirshahvalad", "Atieh", "" ], [ "Lindholm", "Johan", "" ], [ "Derlen", "Mattias", "" ], [ "Rosvall", "Martin", "" ] ]
TITLE: Significant communities in large sparse networks ABSTRACT: Researchers use community-detection algorithms to reveal large-scale organization in biological and social networks, but community detection is useful only if the communities are significant and not a result of noisy data. To assess the statistical significance of the network communities, or the robustness of the detected structure, one approach is to perturb the network structure by removing links and measure how much the communities change. However, perturbing sparse networks is challenging because they are inherently sensitive; they shatter easily if links are removed. Here we propose a simple method to perturb sparse networks and assess the significance of their communities. We generate resampled networks by adding extra links based on local information, then we aggregate the information from multiple resampled networks to find a coarse-grained description of significant clusters. In addition to testing our method on benchmark networks, we use our method on the sparse network of the European Court of Justice (ECJ) case law, to detect significant and insignificant areas of law. We use our significance analysis to draw a map of the ECJ case law network that reveals the relations between the areas of law.
no_new_dataset
0.709466
1208.4586
Jeremiah Blocki
Jeremiah Blocki, Avrim Blum, Anupam Datta and Or Sheffet
Differentially Private Data Analysis of Social Networks via Restricted Sensitivity
null
null
null
null
cs.CR cs.SI physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce the notion of restricted sensitivity as an alternative to global and smooth sensitivity to improve accuracy in differentially private data analysis. The definition of restricted sensitivity is similar to that of global sensitivity except that instead of quantifying over all possible datasets, we take advantage of any beliefs about the dataset that a querier may have, to quantify over a restricted class of datasets. Specifically, given a query f and a hypothesis H about the structure of a dataset D, we show generically how to transform f into a new query f_H whose global sensitivity (over all datasets including those that do not satisfy H) matches the restricted sensitivity of the query f. Moreover, if the belief of the querier is correct (i.e., D is in H) then f_H(D) = f(D). If the belief is incorrect, then f_H(D) may be inaccurate. We demonstrate the usefulness of this notion by considering the task of answering queries regarding social-networks, which we model as a combination of a graph and a labeling of its vertices. In particular, while our generic procedure is computationally inefficient, for the specific definition of H as graphs of bounded degree, we exhibit efficient ways of constructing f_H using different projection-based techniques. We then analyze two important query classes: subgraph counting queries (e.g., number of triangles) and local profile queries (e.g., number of people who know a spy and a computer-scientist who know each other). We demonstrate that the restricted sensitivity of such queries can be significantly lower than their smooth sensitivity. Thus, using restricted sensitivity we can maintain privacy whether or not D is in H, while providing more accurate results in the event that H holds true.
[ { "version": "v1", "created": "Wed, 22 Aug 2012 19:31:05 GMT" }, { "version": "v2", "created": "Fri, 1 Feb 2013 20:33:05 GMT" } ]
2013-02-04T00:00:00
[ [ "Blocki", "Jeremiah", "" ], [ "Blum", "Avrim", "" ], [ "Datta", "Anupam", "" ], [ "Sheffet", "Or", "" ] ]
TITLE: Differentially Private Data Analysis of Social Networks via Restricted Sensitivity ABSTRACT: We introduce the notion of restricted sensitivity as an alternative to global and smooth sensitivity to improve accuracy in differentially private data analysis. The definition of restricted sensitivity is similar to that of global sensitivity except that instead of quantifying over all possible datasets, we take advantage of any beliefs about the dataset that a querier may have, to quantify over a restricted class of datasets. Specifically, given a query f and a hypothesis H about the structure of a dataset D, we show generically how to transform f into a new query f_H whose global sensitivity (over all datasets including those that do not satisfy H) matches the restricted sensitivity of the query f. Moreover, if the belief of the querier is correct (i.e., D is in H) then f_H(D) = f(D). If the belief is incorrect, then f_H(D) may be inaccurate. We demonstrate the usefulness of this notion by considering the task of answering queries regarding social-networks, which we model as a combination of a graph and a labeling of its vertices. In particular, while our generic procedure is computationally inefficient, for the specific definition of H as graphs of bounded degree, we exhibit efficient ways of constructing f_H using different projection-based techniques. We then analyze two important query classes: subgraph counting queries (e.g., number of triangles) and local profile queries (e.g., number of people who know a spy and a computer-scientist who know each other). We demonstrate that the restricted sensitivity of such queries can be significantly lower than their smooth sensitivity. Thus, using restricted sensitivity we can maintain privacy whether or not D is in H, while providing more accurate results in the event that H holds true.
no_new_dataset
0.708811
1210.5338
Cyril Furtlehner
Cyril Furtlehner, Yufei Han, Jean-Marc Lasgouttes and Victorin Martin
Pairwise MRF Calibration by Perturbation of the Bethe Reference Point
54 pages, 8 figure. section 5 and refs added in V2
null
null
Inria RR-8059
cond-mat.dis-nn cond-mat.stat-mech cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We investigate different ways of generating approximate solutions to the pairwise Markov random field (MRF) selection problem. We focus mainly on the inverse Ising problem, but discuss also the somewhat related inverse Gaussian problem because both types of MRF are suitable for inference tasks with the belief propagation algorithm (BP) under certain conditions. Our approach consists in to take a Bethe mean-field solution obtained with a maximum spanning tree (MST) of pairwise mutual information, referred to as the \emph{Bethe reference point}, for further perturbation procedures. We consider three different ways following this idea: in the first one, we select and calibrate iteratively the optimal links to be added starting from the Bethe reference point; the second one is based on the observation that the natural gradient can be computed analytically at the Bethe point; in the third one, assuming no local field and using low temperature expansion we develop a dual loop joint model based on a well chosen fundamental cycle basis. We indeed identify a subclass of planar models, which we refer to as \emph{Bethe-dual graph models}, having possibly many loops, but characterized by a singly connected dual factor graph, for which the partition function and the linear response can be computed exactly in respectively O(N) and $O(N^2)$ operations, thanks to a dual weight propagation (DWP) message passing procedure that we set up. When restricted to this subclass of models, the inverse Ising problem being convex, becomes tractable at any temperature. Experimental tests on various datasets with refined $L_0$ or $L_1$ regularization procedures indicate that these approaches may be competitive and useful alternatives to existing ones.
[ { "version": "v1", "created": "Fri, 19 Oct 2012 08:08:55 GMT" }, { "version": "v2", "created": "Fri, 1 Feb 2013 17:32:44 GMT" } ]
2013-02-04T00:00:00
[ [ "Furtlehner", "Cyril", "" ], [ "Han", "Yufei", "" ], [ "Lasgouttes", "Jean-Marc", "" ], [ "Martin", "Victorin", "" ] ]
TITLE: Pairwise MRF Calibration by Perturbation of the Bethe Reference Point ABSTRACT: We investigate different ways of generating approximate solutions to the pairwise Markov random field (MRF) selection problem. We focus mainly on the inverse Ising problem, but discuss also the somewhat related inverse Gaussian problem because both types of MRF are suitable for inference tasks with the belief propagation algorithm (BP) under certain conditions. Our approach consists in to take a Bethe mean-field solution obtained with a maximum spanning tree (MST) of pairwise mutual information, referred to as the \emph{Bethe reference point}, for further perturbation procedures. We consider three different ways following this idea: in the first one, we select and calibrate iteratively the optimal links to be added starting from the Bethe reference point; the second one is based on the observation that the natural gradient can be computed analytically at the Bethe point; in the third one, assuming no local field and using low temperature expansion we develop a dual loop joint model based on a well chosen fundamental cycle basis. We indeed identify a subclass of planar models, which we refer to as \emph{Bethe-dual graph models}, having possibly many loops, but characterized by a singly connected dual factor graph, for which the partition function and the linear response can be computed exactly in respectively O(N) and $O(N^2)$ operations, thanks to a dual weight propagation (DWP) message passing procedure that we set up. When restricted to this subclass of models, the inverse Ising problem being convex, becomes tractable at any temperature. Experimental tests on various datasets with refined $L_0$ or $L_1$ regularization procedures indicate that these approaches may be competitive and useful alternatives to existing ones.
no_new_dataset
0.709233
1111.5534
Lazaros Gallos
Lazaros K. Gallos, Diego Rybski, Fredrik Liljeros, Shlomo Havlin, Hernan A. Makse
How people interact in evolving online affiliation networks
10 pages, 8 figures
Phys. Rev. X 2, 031014 (2012)
null
null
physics.soc-ph cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The study of human interactions is of central importance for understanding the behavior of individuals, groups and societies. Here, we observe the formation and evolution of networks by monitoring the addition of all new links and we analyze quantitatively the tendencies used to create ties in these evolving online affiliation networks. We first show that an accurate estimation of these probabilistic tendencies can only be achieved by following the time evolution of the network. For example, actions that are attributed to the usual friend of a friend mechanism through a static snapshot of the network are overestimated by a factor of two. A detailed analysis of the dynamic network evolution shows that half of those triangles were generated through other mechanisms, in spite of the characteristic static pattern. We start by characterizing every single link when the tie was established in the network. This allows us to describe the probabilistic tendencies of tie formation and extract sociological conclusions as follows. The tendencies to add new links differ significantly from what we would expect if they were not affected by the individuals' structural position in the network, i.e., from random link formation. We also find significant differences in behavioral traits among individuals according to their degree of activity, gender, age, popularity and other attributes. For instance, in the particular datasets analyzed here, we find that women reciprocate connections three times as much as men and this difference increases with age. Men tend to connect with the most popular people more often than women across all ages. On the other hand, triangular ties tendencies are similar and independent of gender. Our findings can be useful to build models of realistic social network structures and discover the underlying laws that govern establishment of ties in evolving social networks.
[ { "version": "v1", "created": "Wed, 23 Nov 2011 16:04:06 GMT" } ]
2013-02-01T00:00:00
[ [ "Gallos", "Lazaros K.", "" ], [ "Rybski", "Diego", "" ], [ "Liljeros", "Fredrik", "" ], [ "Havlin", "Shlomo", "" ], [ "Makse", "Hernan A.", "" ] ]
TITLE: How people interact in evolving online affiliation networks ABSTRACT: The study of human interactions is of central importance for understanding the behavior of individuals, groups and societies. Here, we observe the formation and evolution of networks by monitoring the addition of all new links and we analyze quantitatively the tendencies used to create ties in these evolving online affiliation networks. We first show that an accurate estimation of these probabilistic tendencies can only be achieved by following the time evolution of the network. For example, actions that are attributed to the usual friend of a friend mechanism through a static snapshot of the network are overestimated by a factor of two. A detailed analysis of the dynamic network evolution shows that half of those triangles were generated through other mechanisms, in spite of the characteristic static pattern. We start by characterizing every single link when the tie was established in the network. This allows us to describe the probabilistic tendencies of tie formation and extract sociological conclusions as follows. The tendencies to add new links differ significantly from what we would expect if they were not affected by the individuals' structural position in the network, i.e., from random link formation. We also find significant differences in behavioral traits among individuals according to their degree of activity, gender, age, popularity and other attributes. For instance, in the particular datasets analyzed here, we find that women reciprocate connections three times as much as men and this difference increases with age. Men tend to connect with the most popular people more often than women across all ages. On the other hand, triangular ties tendencies are similar and independent of gender. Our findings can be useful to build models of realistic social network structures and discover the underlying laws that govern establishment of ties in evolving social networks.
no_new_dataset
0.705969
1301.7363
John S. Breese
John S. Breese, David Heckerman, Carl Kadie
Empirical Analysis of Predictive Algorithms for Collaborative Filtering
Appears in Proceedings of the Fourteenth Conference on Uncertainty in Artificial Intelligence (UAI1998)
null
null
UAI-P-1998-PG-43-52
cs.IR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Collaborative filtering or recommender systems use a database about user preferences to predict additional topics or products a new user might like. In this paper we describe several algorithms designed for this task, including techniques based on correlation coefficients, vector-based similarity calculations, and statistical Bayesian methods. We compare the predictive accuracy of the various methods in a set of representative problem domains. We use two basic classes of evaluation metrics. The first characterizes accuracy over a set of individual predictions in terms of average absolute deviation. The second estimates the utility of a ranked list of suggested items. This metric uses an estimate of the probability that a user will see a recommendation in an ordered list. Experiments were run for datasets associated with 3 application areas, 4 experimental protocols, and the 2 evaluation metrics for the various algorithms. Results indicate that for a wide range of conditions, Bayesian networks with decision trees at each node and correlation methods outperform Bayesian-clustering and vector-similarity methods. Between correlation and Bayesian networks, the preferred method depends on the nature of the dataset, nature of the application (ranked versus one-by-one presentation), and the availability of votes with which to make predictions. Other considerations include the size of database, speed of predictions, and learning time.
[ { "version": "v1", "created": "Wed, 30 Jan 2013 15:02:44 GMT" } ]
2013-02-01T00:00:00
[ [ "Breese", "John S.", "" ], [ "Heckerman", "David", "" ], [ "Kadie", "Carl", "" ] ]
TITLE: Empirical Analysis of Predictive Algorithms for Collaborative Filtering ABSTRACT: Collaborative filtering or recommender systems use a database about user preferences to predict additional topics or products a new user might like. In this paper we describe several algorithms designed for this task, including techniques based on correlation coefficients, vector-based similarity calculations, and statistical Bayesian methods. We compare the predictive accuracy of the various methods in a set of representative problem domains. We use two basic classes of evaluation metrics. The first characterizes accuracy over a set of individual predictions in terms of average absolute deviation. The second estimates the utility of a ranked list of suggested items. This metric uses an estimate of the probability that a user will see a recommendation in an ordered list. Experiments were run for datasets associated with 3 application areas, 4 experimental protocols, and the 2 evaluation metrics for the various algorithms. Results indicate that for a wide range of conditions, Bayesian networks with decision trees at each node and correlation methods outperform Bayesian-clustering and vector-similarity methods. Between correlation and Bayesian networks, the preferred method depends on the nature of the dataset, nature of the application (ranked versus one-by-one presentation), and the availability of votes with which to make predictions. Other considerations include the size of database, speed of predictions, and learning time.
no_new_dataset
0.711982
1301.7676
Nicos Angelopoulos
Anthony Monnet and Roger Villemaire
Efficient Partial Order CDCL Using Assertion Level Choice Heuristics
Appeared in CICLOPS 2012. 15 Pages, 1 Figure
null
null
null
cs.AI cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We previously designed Partial Order Conflict Driven Clause Learning (PO-CDCL), a variation of the satisfiability solving CDCL algorithm with a partial order on decision levels, and showed that it can speed up the solving on problems with a high independence between decision levels. In this paper, we more thoroughly analyze the reasons of the efficiency of PO-CDCL. Of particular importance is that the partial order introduces several candidates for the assertion level. By evaluating different heuristics for this choice, we show that the assertion level selection has an important impact on solving and that a carefully designed heuristic can significantly improve performances on relevant benchmarks.
[ { "version": "v1", "created": "Thu, 31 Jan 2013 16:45:01 GMT" } ]
2013-02-01T00:00:00
[ [ "Monnet", "Anthony", "" ], [ "Villemaire", "Roger", "" ] ]
TITLE: Efficient Partial Order CDCL Using Assertion Level Choice Heuristics ABSTRACT: We previously designed Partial Order Conflict Driven Clause Learning (PO-CDCL), a variation of the satisfiability solving CDCL algorithm with a partial order on decision levels, and showed that it can speed up the solving on problems with a high independence between decision levels. In this paper, we more thoroughly analyze the reasons of the efficiency of PO-CDCL. Of particular importance is that the partial order introduces several candidates for the assertion level. By evaluating different heuristics for this choice, we show that the assertion level selection has an important impact on solving and that a carefully designed heuristic can significantly improve performances on relevant benchmarks.
no_new_dataset
0.711017
0807.1494
Matteo Gagliolo
Matteo Gagliolo and Juergen Schmidhuber
Algorithm Selection as a Bandit Problem with Unbounded Losses
15 pages, 2 figures
null
10.1007/978-3-642-13800-3_7
IDSIA-07-08
cs.AI cs.GT cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Algorithm selection is typically based on models of algorithm performance, learned during a separate offline training sequence, which can be prohibitively expensive. In recent work, we adopted an online approach, in which a performance model is iteratively updated and used to guide selection on a sequence of problem instances. The resulting exploration-exploitation trade-off was represented as a bandit problem with expert advice, using an existing solver for this game, but this required the setting of an arbitrary bound on algorithm runtimes, thus invalidating the optimal regret of the solver. In this paper, we propose a simpler framework for representing algorithm selection as a bandit problem, with partial information, and an unknown bound on losses. We adapt an existing solver to this game, proving a bound on its expected regret, which holds also for the resulting algorithm selection technique. We present preliminary experiments with a set of SAT solvers on a mixed SAT-UNSAT benchmark.
[ { "version": "v1", "created": "Wed, 9 Jul 2008 16:47:36 GMT" } ]
2013-01-31T00:00:00
[ [ "Gagliolo", "Matteo", "" ], [ "Schmidhuber", "Juergen", "" ] ]
TITLE: Algorithm Selection as a Bandit Problem with Unbounded Losses ABSTRACT: Algorithm selection is typically based on models of algorithm performance, learned during a separate offline training sequence, which can be prohibitively expensive. In recent work, we adopted an online approach, in which a performance model is iteratively updated and used to guide selection on a sequence of problem instances. The resulting exploration-exploitation trade-off was represented as a bandit problem with expert advice, using an existing solver for this game, but this required the setting of an arbitrary bound on algorithm runtimes, thus invalidating the optimal regret of the solver. In this paper, we propose a simpler framework for representing algorithm selection as a bandit problem, with partial information, and an unknown bound on losses. We adapt an existing solver to this game, proving a bound on its expected regret, which holds also for the resulting algorithm selection technique. We present preliminary experiments with a set of SAT solvers on a mixed SAT-UNSAT benchmark.
no_new_dataset
0.710208
1202.3861
Michael Schreiber
Michael Schreiber
Inconsistencies of Recently Proposed Citation Impact Indicators and how to Avoid Them
14 pages, 9 figures, accepted by Journal of the American Society for Information Science and Technology Final version with slightly changed figures, new scoring rule, extended discussion
J. Am. Soc. Inf. Sci. Techn. 63(10), 2062-2073, (2012)
null
null
stat.AP cs.DL physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
It is shown that under certain circumstances in particular for small datasets the recently proposed citation impact indicators I3(6PR) and R(6,k) behave inconsistently when additional papers or citations are taken into consideration. Three simple examples are presented, in which the indicators fluctuate strongly and the ranking of scientists in the evaluated group is sometimes completely mixed up by minor changes in the data base. The erratic behavior is traced to the specific way in which weights are attributed to the six percentile rank classes, specifically for the tied papers. For 100 percentile rank classes the effects will be less serious. For the 6 classes it is demonstrated that a different way of assigning weights avoids these problems, although the non-linearity of the weights for the different percentile rank classes can still lead to (much less frequent) changes in the ranking. This behavior is not undesired, because it can be used to correct for differences in citation behavior in different fields. Remaining deviations from the theoretical value R(6,k) = 1.91 can be avoided by a new scoring rule, the fractional scoring. Previously proposed consistency criteria are amended by another property of strict independence which a performance indicator should aim at.
[ { "version": "v1", "created": "Fri, 17 Feb 2012 10:05:04 GMT" }, { "version": "v2", "created": "Wed, 4 Apr 2012 08:33:52 GMT" } ]
2013-01-31T00:00:00
[ [ "Schreiber", "Michael", "" ] ]
TITLE: Inconsistencies of Recently Proposed Citation Impact Indicators and how to Avoid Them ABSTRACT: It is shown that under certain circumstances in particular for small datasets the recently proposed citation impact indicators I3(6PR) and R(6,k) behave inconsistently when additional papers or citations are taken into consideration. Three simple examples are presented, in which the indicators fluctuate strongly and the ranking of scientists in the evaluated group is sometimes completely mixed up by minor changes in the data base. The erratic behavior is traced to the specific way in which weights are attributed to the six percentile rank classes, specifically for the tied papers. For 100 percentile rank classes the effects will be less serious. For the 6 classes it is demonstrated that a different way of assigning weights avoids these problems, although the non-linearity of the weights for the different percentile rank classes can still lead to (much less frequent) changes in the ranking. This behavior is not undesired, because it can be used to correct for differences in citation behavior in different fields. Remaining deviations from the theoretical value R(6,k) = 1.91 can be avoided by a new scoring rule, the fractional scoring. Previously proposed consistency criteria are amended by another property of strict independence which a performance indicator should aim at.
no_new_dataset
0.710622
1211.2756
Anton Korobeynikov
Sergey I. Nikolenko, Anton I. Korobeynikov and Max A. Alekseyev
BayesHammer: Bayesian clustering for error correction in single-cell sequencing
null
BMC Genomics 14(Suppl 1) (2013), pp. S7
10.1186/1471-2164-14-S1-S7
null
q-bio.QM cs.CE cs.DS q-bio.GN
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Error correction of sequenced reads remains a difficult task, especially in single-cell sequencing projects with extremely non-uniform coverage. While existing error correction tools designed for standard (multi-cell) sequencing data usually come up short in single-cell sequencing projects, algorithms actually used for single-cell error correction have been so far very simplistic. We introduce several novel algorithms based on Hamming graphs and Bayesian subclustering in our new error correction tool BayesHammer. While BayesHammer was designed for single-cell sequencing, we demonstrate that it also improves on existing error correction tools for multi-cell sequencing data while working much faster on real-life datasets. We benchmark BayesHammer on both $k$-mer counts and actual assembly results with the SPAdes genome assembler.
[ { "version": "v1", "created": "Mon, 12 Nov 2012 19:52:34 GMT" } ]
2013-01-31T00:00:00
[ [ "Nikolenko", "Sergey I.", "" ], [ "Korobeynikov", "Anton I.", "" ], [ "Alekseyev", "Max A.", "" ] ]
TITLE: BayesHammer: Bayesian clustering for error correction in single-cell sequencing ABSTRACT: Error correction of sequenced reads remains a difficult task, especially in single-cell sequencing projects with extremely non-uniform coverage. While existing error correction tools designed for standard (multi-cell) sequencing data usually come up short in single-cell sequencing projects, algorithms actually used for single-cell error correction have been so far very simplistic. We introduce several novel algorithms based on Hamming graphs and Bayesian subclustering in our new error correction tool BayesHammer. While BayesHammer was designed for single-cell sequencing, we demonstrate that it also improves on existing error correction tools for multi-cell sequencing data while working much faster on real-life datasets. We benchmark BayesHammer on both $k$-mer counts and actual assembly results with the SPAdes genome assembler.
no_new_dataset
0.709044
1301.6939
Edward Grefenstette
Edward Grefenstette, Georgiana Dinu, Yao-Zhong Zhang, Mehrnoosh Sadrzadeh and Marco Baroni
Multi-Step Regression Learning for Compositional Distributional Semantics
10 pages + 1 page references, to be presented at the 10th International Conference on Computational Semantics (IWCS 2013)
null
null
null
cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a model for compositional distributional semantics related to the framework of Coecke et al. (2010), and emulating formal semantics by representing functions as tensors and arguments as vectors. We introduce a new learning method for tensors, generalising the approach of Baroni and Zamparelli (2010). We evaluate it on two benchmark data sets, and find it to outperform existing leading methods. We argue in our analysis that the nature of this learning method also renders it suitable for solving more subtle problems compositional distributional models might face.
[ { "version": "v1", "created": "Tue, 29 Jan 2013 14:59:34 GMT" }, { "version": "v2", "created": "Wed, 30 Jan 2013 12:01:23 GMT" } ]
2013-01-31T00:00:00
[ [ "Grefenstette", "Edward", "" ], [ "Dinu", "Georgiana", "" ], [ "Zhang", "Yao-Zhong", "" ], [ "Sadrzadeh", "Mehrnoosh", "" ], [ "Baroni", "Marco", "" ] ]
TITLE: Multi-Step Regression Learning for Compositional Distributional Semantics ABSTRACT: We present a model for compositional distributional semantics related to the framework of Coecke et al. (2010), and emulating formal semantics by representing functions as tensors and arguments as vectors. We introduce a new learning method for tensors, generalising the approach of Baroni and Zamparelli (2010). We evaluate it on two benchmark data sets, and find it to outperform existing leading methods. We argue in our analysis that the nature of this learning method also renders it suitable for solving more subtle problems compositional distributional models might face.
no_new_dataset
0.710176
0910.5260
Sewoong Oh
Raghunandan H. Keshavan, Sewoong Oh
A Gradient Descent Algorithm on the Grassman Manifold for Matrix Completion
26 pages, 15 figures
null
10.1016/j.trc.2012.12.007
null
cs.NA cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the problem of reconstructing a low-rank matrix from a small subset of its entries. In this paper, we describe the implementation of an efficient algorithm called OptSpace, based on singular value decomposition followed by local manifold optimization, for solving the low-rank matrix completion problem. It has been shown that if the number of revealed entries is large enough, the output of singular value decomposition gives a good estimate for the original matrix, so that local optimization reconstructs the correct matrix with high probability. We present numerical results which show that this algorithm can reconstruct the low rank matrix exactly from a very small subset of its entries. We further study the robustness of the algorithm with respect to noise, and its performance on actual collaborative filtering datasets.
[ { "version": "v1", "created": "Tue, 27 Oct 2009 22:19:31 GMT" }, { "version": "v2", "created": "Tue, 3 Nov 2009 23:35:13 GMT" } ]
2013-01-30T00:00:00
[ [ "Keshavan", "Raghunandan H.", "" ], [ "Oh", "Sewoong", "" ] ]
TITLE: A Gradient Descent Algorithm on the Grassman Manifold for Matrix Completion ABSTRACT: We consider the problem of reconstructing a low-rank matrix from a small subset of its entries. In this paper, we describe the implementation of an efficient algorithm called OptSpace, based on singular value decomposition followed by local manifold optimization, for solving the low-rank matrix completion problem. It has been shown that if the number of revealed entries is large enough, the output of singular value decomposition gives a good estimate for the original matrix, so that local optimization reconstructs the correct matrix with high probability. We present numerical results which show that this algorithm can reconstruct the low rank matrix exactly from a very small subset of its entries. We further study the robustness of the algorithm with respect to noise, and its performance on actual collaborative filtering datasets.
no_new_dataset
0.711268
1205.4377
Kirill Trapeznikov
Kirill Trapeznikov, Venkatesh Saligrama, David Castanon
Multi-Stage Classifier Design
null
null
null
null
cs.CV stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In many classification systems, sensing modalities have different acquisition costs. It is often {\it unnecessary} to use every modality to classify a majority of examples. We study a multi-stage system in a prediction time cost reduction setting, where the full data is available for training, but for a test example, measurements in a new modality can be acquired at each stage for an additional cost. We seek decision rules to reduce the average measurement acquisition cost. We formulate an empirical risk minimization problem (ERM) for a multi-stage reject classifier, wherein the stage $k$ classifier either classifies a sample using only the measurements acquired so far or rejects it to the next stage where more attributes can be acquired for a cost. To solve the ERM problem, we show that the optimal reject classifier at each stage is a combination of two binary classifiers, one biased towards positive examples and the other biased towards negative examples. We use this parameterization to construct stage-by-stage global surrogate risk, develop an iterative algorithm in the boosting framework and present convergence and generalization results. We test our work on synthetic, medical and explosives detection datasets. Our results demonstrate that substantial cost reduction without a significant sacrifice in accuracy is achievable.
[ { "version": "v1", "created": "Sun, 20 May 2012 03:15:13 GMT" }, { "version": "v2", "created": "Tue, 29 Jan 2013 16:54:30 GMT" } ]
2013-01-30T00:00:00
[ [ "Trapeznikov", "Kirill", "" ], [ "Saligrama", "Venkatesh", "" ], [ "Castanon", "David", "" ] ]
TITLE: Multi-Stage Classifier Design ABSTRACT: In many classification systems, sensing modalities have different acquisition costs. It is often {\it unnecessary} to use every modality to classify a majority of examples. We study a multi-stage system in a prediction time cost reduction setting, where the full data is available for training, but for a test example, measurements in a new modality can be acquired at each stage for an additional cost. We seek decision rules to reduce the average measurement acquisition cost. We formulate an empirical risk minimization problem (ERM) for a multi-stage reject classifier, wherein the stage $k$ classifier either classifies a sample using only the measurements acquired so far or rejects it to the next stage where more attributes can be acquired for a cost. To solve the ERM problem, we show that the optimal reject classifier at each stage is a combination of two binary classifiers, one biased towards positive examples and the other biased towards negative examples. We use this parameterization to construct stage-by-stage global surrogate risk, develop an iterative algorithm in the boosting framework and present convergence and generalization results. We test our work on synthetic, medical and explosives detection datasets. Our results demonstrate that substantial cost reduction without a significant sacrifice in accuracy is achievable.
no_new_dataset
0.709856
1301.6686
Gregory F. Cooper
Gregory F. Cooper, Changwon Yoo
Causal Discovery from a Mixture of Experimental and Observational Data
Appears in Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI1999)
null
null
UAI-P-1999-PG-116-125
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper describes a Bayesian method for combining an arbitrary mixture of observational and experimental data in order to learn causal Bayesian networks. Observational data are passively observed. Experimental data, such as that produced by randomized controlled trials, result from the experimenter manipulating one or more variables (typically randomly) and observing the states of other variables. The paper presents a Bayesian method for learning the causal structure and parameters of the underlying causal process that is generating the data, given that (1) the data contains a mixture of observational and experimental case records, and (2) the causal process is modeled as a causal Bayesian network. This learning method was applied using as input various mixtures of experimental and observational data that were generated from the ALARM causal Bayesian network. In these experiments, the absolute and relative quantities of experimental and observational data were varied systematically. For each of these training datasets, the learning method was applied to predict the causal structure and to estimate the causal parameters that exist among randomly selected pairs of nodes in ALARM that are not confounded. The paper reports how these structure predictions and parameter estimates compare with the true causal structures and parameters as given by the ALARM network.
[ { "version": "v1", "created": "Wed, 23 Jan 2013 15:57:22 GMT" } ]
2013-01-30T00:00:00
[ [ "Cooper", "Gregory F.", "" ], [ "Yoo", "Changwon", "" ] ]
TITLE: Causal Discovery from a Mixture of Experimental and Observational Data ABSTRACT: This paper describes a Bayesian method for combining an arbitrary mixture of observational and experimental data in order to learn causal Bayesian networks. Observational data are passively observed. Experimental data, such as that produced by randomized controlled trials, result from the experimenter manipulating one or more variables (typically randomly) and observing the states of other variables. The paper presents a Bayesian method for learning the causal structure and parameters of the underlying causal process that is generating the data, given that (1) the data contains a mixture of observational and experimental case records, and (2) the causal process is modeled as a causal Bayesian network. This learning method was applied using as input various mixtures of experimental and observational data that were generated from the ALARM causal Bayesian network. In these experiments, the absolute and relative quantities of experimental and observational data were varied systematically. For each of these training datasets, the learning method was applied to predict the causal structure and to estimate the causal parameters that exist among randomly selected pairs of nodes in ALARM that are not confounded. The paper reports how these structure predictions and parameter estimates compare with the true causal structures and parameters as given by the ALARM network.
no_new_dataset
0.712607
1301.6723
Stefano Monti
Stefano Monti, Gregory F. Cooper
A Bayesian Network Classifier that Combines a Finite Mixture Model and a Naive Bayes Model
Appears in Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence (UAI1999)
null
null
UAI-P-1999-PG-447-456
cs.LG cs.AI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we present a new Bayesian network model for classification that combines the naive-Bayes (NB) classifier and the finite-mixture (FM) classifier. The resulting classifier aims at relaxing the strong assumptions on which the two component models are based, in an attempt to improve on their classification performance, both in terms of accuracy and in terms of calibration of the estimated probabilities. The proposed classifier is obtained by superimposing a finite mixture model on the set of feature variables of a naive Bayes model. We present experimental results that compare the predictive performance on real datasets of the new classifier with the predictive performance of the NB classifier and the FM classifier.
[ { "version": "v1", "created": "Wed, 23 Jan 2013 15:59:54 GMT" } ]
2013-01-30T00:00:00
[ [ "Monti", "Stefano", "" ], [ "Cooper", "Gregory F.", "" ] ]
TITLE: A Bayesian Network Classifier that Combines a Finite Mixture Model and a Naive Bayes Model ABSTRACT: In this paper we present a new Bayesian network model for classification that combines the naive-Bayes (NB) classifier and the finite-mixture (FM) classifier. The resulting classifier aims at relaxing the strong assumptions on which the two component models are based, in an attempt to improve on their classification performance, both in terms of accuracy and in terms of calibration of the estimated probabilities. The proposed classifier is obtained by superimposing a finite mixture model on the set of feature variables of a naive Bayes model. We present experimental results that compare the predictive performance on real datasets of the new classifier with the predictive performance of the NB classifier and the FM classifier.
no_new_dataset
0.713251
1301.6770
Zhixiang Eddie Xu
Zhixiang (Eddie) Xu, Minmin Chen, Kilian Q. Weinberger, Fei Sha
An alternative text representation to TF-IDF and Bag-of-Words
null
null
null
null
cs.IR cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In text mining, information retrieval, and machine learning, text documents are commonly represented through variants of sparse Bag of Words (sBoW) vectors (e.g. TF-IDF). Although simple and intuitive, sBoW style representations suffer from their inherent over-sparsity and fail to capture word-level synonymy and polysemy. Especially when labeled data is limited (e.g. in document classification), or the text documents are short (e.g. emails or abstracts), many features are rarely observed within the training corpus. This leads to overfitting and reduced generalization accuracy. In this paper we propose Dense Cohort of Terms (dCoT), an unsupervised algorithm to learn improved sBoW document features. dCoT explicitly models absent words by removing and reconstructing random sub-sets of words in the unlabeled corpus. With this approach, dCoT learns to reconstruct frequent words from co-occurring infrequent words and maps the high dimensional sparse sBoW vectors into a low-dimensional dense representation. We show that the feature removal can be marginalized out and that the reconstruction can be solved for in closed-form. We demonstrate empirically, on several benchmark datasets, that dCoT features significantly improve the classification accuracy across several document classification tasks.
[ { "version": "v1", "created": "Mon, 28 Jan 2013 21:04:45 GMT" } ]
2013-01-30T00:00:00
[ [ "Zhixiang", "", "", "Eddie" ], [ "Xu", "", "" ], [ "Chen", "Minmin", "" ], [ "Weinberger", "Kilian Q.", "" ], [ "Sha", "Fei", "" ] ]
TITLE: An alternative text representation to TF-IDF and Bag-of-Words ABSTRACT: In text mining, information retrieval, and machine learning, text documents are commonly represented through variants of sparse Bag of Words (sBoW) vectors (e.g. TF-IDF). Although simple and intuitive, sBoW style representations suffer from their inherent over-sparsity and fail to capture word-level synonymy and polysemy. Especially when labeled data is limited (e.g. in document classification), or the text documents are short (e.g. emails or abstracts), many features are rarely observed within the training corpus. This leads to overfitting and reduced generalization accuracy. In this paper we propose Dense Cohort of Terms (dCoT), an unsupervised algorithm to learn improved sBoW document features. dCoT explicitly models absent words by removing and reconstructing random sub-sets of words in the unlabeled corpus. With this approach, dCoT learns to reconstruct frequent words from co-occurring infrequent words and maps the high dimensional sparse sBoW vectors into a low-dimensional dense representation. We show that the feature removal can be marginalized out and that the reconstruction can be solved for in closed-form. We demonstrate empirically, on several benchmark datasets, that dCoT features significantly improve the classification accuracy across several document classification tasks.
no_new_dataset
0.709868
1301.6870
Paridhi Jain
Anshu Malhotra, Luam Totti, Wagner Meira Jr., Ponnurangam Kumaraguru, Virgilio Almeida
Studying User Footprints in Different Online Social Networks
The paper is already published in ASONAM 2012
null
null
null
cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the growing popularity and usage of online social media services, people now have accounts (some times several) on multiple and diverse services like Facebook, LinkedIn, Twitter and YouTube. Publicly available information can be used to create a digital footprint of any user using these social media services. Generating such digital footprints can be very useful for personalization, profile management, detecting malicious behavior of users. A very important application of analyzing users' online digital footprints is to protect users from potential privacy and security risks arising from the huge publicly available user information. We extracted information about user identities on different social networks through Social Graph API, FriendFeed, and Profilactic; we collated our own dataset to create the digital footprints of the users. We used username, display name, description, location, profile image, and number of connections to generate the digital footprints of the user. We applied context specific techniques (e.g. Jaro Winkler similarity, Wordnet based ontologies) to measure the similarity of the user profiles on different social networks. We specifically focused on Twitter and LinkedIn. In this paper, we present the analysis and results from applying automated classifiers for disambiguating profiles belonging to the same user from different social networks. UserID and Name were found to be the most discriminative features for disambiguating user profiles. Using the most promising set of features and similarity metrics, we achieved accuracy, precision and recall of 98%, 99%, and 96%, respectively.
[ { "version": "v1", "created": "Tue, 29 Jan 2013 09:29:54 GMT" } ]
2013-01-30T00:00:00
[ [ "Malhotra", "Anshu", "" ], [ "Totti", "Luam", "" ], [ "Meira", "Wagner", "Jr." ], [ "Kumaraguru", "Ponnurangam", "" ], [ "Almeida", "Virgilio", "" ] ]
TITLE: Studying User Footprints in Different Online Social Networks ABSTRACT: With the growing popularity and usage of online social media services, people now have accounts (some times several) on multiple and diverse services like Facebook, LinkedIn, Twitter and YouTube. Publicly available information can be used to create a digital footprint of any user using these social media services. Generating such digital footprints can be very useful for personalization, profile management, detecting malicious behavior of users. A very important application of analyzing users' online digital footprints is to protect users from potential privacy and security risks arising from the huge publicly available user information. We extracted information about user identities on different social networks through Social Graph API, FriendFeed, and Profilactic; we collated our own dataset to create the digital footprints of the users. We used username, display name, description, location, profile image, and number of connections to generate the digital footprints of the user. We applied context specific techniques (e.g. Jaro Winkler similarity, Wordnet based ontologies) to measure the similarity of the user profiles on different social networks. We specifically focused on Twitter and LinkedIn. In this paper, we present the analysis and results from applying automated classifiers for disambiguating profiles belonging to the same user from different social networks. UserID and Name were found to be the most discriminative features for disambiguating user profiles. Using the most promising set of features and similarity metrics, we achieved accuracy, precision and recall of 98%, 99%, and 96%, respectively.
new_dataset
0.716993
1210.0137
Pierre Deville Pierre
Vincent D. Blondel, Markus Esch, Connie Chan, Fabrice Clerot, Pierre Deville, Etienne Huens, Fr\'ed\'eric Morlot, Zbigniew Smoreda and Cezary Ziemlicki
Data for Development: the D4D Challenge on Mobile Phone Data
10 pages, 3 figures
null
null
null
cs.CY cs.SI physics.soc-ph stat.CO
http://creativecommons.org/licenses/by-nc-sa/3.0/
The Orange "Data for Development" (D4D) challenge is an open data challenge on anonymous call patterns of Orange's mobile phone users in Ivory Coast. The goal of the challenge is to help address society development questions in novel ways by contributing to the socio-economic development and well-being of the Ivory Coast population. Participants to the challenge are given access to four mobile phone datasets and the purpose of this paper is to describe the four datasets. The website http://www.d4d.orange.com contains more information about the participation rules. The datasets are based on anonymized Call Detail Records (CDR) of phone calls and SMS exchanges between five million of Orange's customers in Ivory Coast between December 1, 2011 and April 28, 2012. The datasets are: (a) antenna-to-antenna traffic on an hourly basis, (b) individual trajectories for 50,000 customers for two week time windows with antenna location information, (3) individual trajectories for 500,000 customers over the entire observation period with sub-prefecture location information, and (4) a sample of communication graphs for 5,000 customers
[ { "version": "v1", "created": "Sat, 29 Sep 2012 17:39:16 GMT" }, { "version": "v2", "created": "Mon, 28 Jan 2013 12:56:55 GMT" } ]
2013-01-29T00:00:00
[ [ "Blondel", "Vincent D.", "" ], [ "Esch", "Markus", "" ], [ "Chan", "Connie", "" ], [ "Clerot", "Fabrice", "" ], [ "Deville", "Pierre", "" ], [ "Huens", "Etienne", "" ], [ "Morlot", "Frédéric", "" ], [ "Smoreda", "Zbigniew", "" ], [ "Ziemlicki", "Cezary", "" ] ]
TITLE: Data for Development: the D4D Challenge on Mobile Phone Data ABSTRACT: The Orange "Data for Development" (D4D) challenge is an open data challenge on anonymous call patterns of Orange's mobile phone users in Ivory Coast. The goal of the challenge is to help address society development questions in novel ways by contributing to the socio-economic development and well-being of the Ivory Coast population. Participants to the challenge are given access to four mobile phone datasets and the purpose of this paper is to describe the four datasets. The website http://www.d4d.orange.com contains more information about the participation rules. The datasets are based on anonymized Call Detail Records (CDR) of phone calls and SMS exchanges between five million of Orange's customers in Ivory Coast between December 1, 2011 and April 28, 2012. The datasets are: (a) antenna-to-antenna traffic on an hourly basis, (b) individual trajectories for 50,000 customers for two week time windows with antenna location information, (3) individual trajectories for 500,000 customers over the entire observation period with sub-prefecture location information, and (4) a sample of communication graphs for 5,000 customers
no_new_dataset
0.69592
1211.4410
Sotirios Chatzis
Emmanouil A. Platanios and Sotirios P. Chatzis
Mixture Gaussian Process Conditional Heteroscedasticity
Technical Report, under preparation
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Generalized autoregressive conditional heteroscedasticity (GARCH) models have long been considered as one of the most successful families of approaches for volatility modeling in financial return series. In this paper, we propose an alternative approach based on methodologies widely used in the field of statistical machine learning. Specifically, we propose a novel nonparametric Bayesian mixture of Gaussian process regression models, each component of which models the noise variance process that contaminates the observed data as a separate latent Gaussian process driven by the observed data. This way, we essentially obtain a mixture Gaussian process conditional heteroscedasticity (MGPCH) model for volatility modeling in financial return series. We impose a nonparametric prior with power-law nature over the distribution of the model mixture components, namely the Pitman-Yor process prior, to allow for better capturing modeled data distributions with heavy tails and skewness. Finally, we provide a copula- based approach for obtaining a predictive posterior for the covariances over the asset returns modeled by means of a postulated MGPCH model. We evaluate the efficacy of our approach in a number of benchmark scenarios, and compare its performance to state-of-the-art methodologies.
[ { "version": "v1", "created": "Mon, 19 Nov 2012 13:33:55 GMT" }, { "version": "v2", "created": "Sun, 2 Dec 2012 20:58:33 GMT" }, { "version": "v3", "created": "Thu, 6 Dec 2012 11:48:29 GMT" }, { "version": "v4", "created": "Fri, 25 Jan 2013 22:03:25 GMT" } ]
2013-01-29T00:00:00
[ [ "Platanios", "Emmanouil A.", "" ], [ "Chatzis", "Sotirios P.", "" ] ]
TITLE: Mixture Gaussian Process Conditional Heteroscedasticity ABSTRACT: Generalized autoregressive conditional heteroscedasticity (GARCH) models have long been considered as one of the most successful families of approaches for volatility modeling in financial return series. In this paper, we propose an alternative approach based on methodologies widely used in the field of statistical machine learning. Specifically, we propose a novel nonparametric Bayesian mixture of Gaussian process regression models, each component of which models the noise variance process that contaminates the observed data as a separate latent Gaussian process driven by the observed data. This way, we essentially obtain a mixture Gaussian process conditional heteroscedasticity (MGPCH) model for volatility modeling in financial return series. We impose a nonparametric prior with power-law nature over the distribution of the model mixture components, namely the Pitman-Yor process prior, to allow for better capturing modeled data distributions with heavy tails and skewness. Finally, we provide a copula- based approach for obtaining a predictive posterior for the covariances over the asset returns modeled by means of a postulated MGPCH model. We evaluate the efficacy of our approach in a number of benchmark scenarios, and compare its performance to state-of-the-art methodologies.
no_new_dataset
0.71103
1301.4293
Limin Yao
Sebastian Riedel, Limin Yao, Andrew McCallum
Latent Relation Representations for Universal Schemas
4 pages, ICLR workshop
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Traditional relation extraction predicts relations within some fixed and finite target schema. Machine learning approaches to this task require either manual annotation or, in the case of distant supervision, existing structured sources of the same schema. The need for existing datasets can be avoided by using a universal schema: the union of all involved schemas (surface form predicates as in OpenIE, and relations in the schemas of pre-existing databases). This schema has an almost unlimited set of relations (due to surface forms), and supports integration with existing structured data (through the relation types of existing databases). To populate a database of such schema we present a family of matrix factorization models that predict affinity between database tuples and relations. We show that this achieves substantially higher accuracy than the traditional classification approach. More importantly, by operating simultaneously on relations observed in text and in pre-existing structured DBs such as Freebase, we are able to reason about unstructured and structured data in mutually-supporting ways. By doing so our approach outperforms state-of-the-art distant supervision systems.
[ { "version": "v1", "created": "Fri, 18 Jan 2013 04:37:30 GMT" }, { "version": "v2", "created": "Mon, 28 Jan 2013 20:10:21 GMT" } ]
2013-01-29T00:00:00
[ [ "Riedel", "Sebastian", "" ], [ "Yao", "Limin", "" ], [ "McCallum", "Andrew", "" ] ]
TITLE: Latent Relation Representations for Universal Schemas ABSTRACT: Traditional relation extraction predicts relations within some fixed and finite target schema. Machine learning approaches to this task require either manual annotation or, in the case of distant supervision, existing structured sources of the same schema. The need for existing datasets can be avoided by using a universal schema: the union of all involved schemas (surface form predicates as in OpenIE, and relations in the schemas of pre-existing databases). This schema has an almost unlimited set of relations (due to surface forms), and supports integration with existing structured data (through the relation types of existing databases). To populate a database of such schema we present a family of matrix factorization models that predict affinity between database tuples and relations. We show that this achieves substantially higher accuracy than the traditional classification approach. More importantly, by operating simultaneously on relations observed in text and in pre-existing structured DBs such as Freebase, we are able to reason about unstructured and structured data in mutually-supporting ways. By doing so our approach outperforms state-of-the-art distant supervision systems.
no_new_dataset
0.708648
1301.5686
Jeon-Hyung Kang
Jeon-Hyung Kang, Jun Ma, Yan Liu
Transfer Topic Modeling with Ease and Scalability
2012 SIAM International Conference on Data Mining (SDM12) Pages: {564-575}
null
null
null
cs.CL cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The increasing volume of short texts generated on social media sites, such as Twitter or Facebook, creates a great demand for effective and efficient topic modeling approaches. While latent Dirichlet allocation (LDA) can be applied, it is not optimal due to its weakness in handling short texts with fast-changing topics and scalability concerns. In this paper, we propose a transfer learning approach that utilizes abundant labeled documents from other domains (such as Yahoo! News or Wikipedia) to improve topic modeling, with better model fitting and result interpretation. Specifically, we develop Transfer Hierarchical LDA (thLDA) model, which incorporates the label information from other domains via informative priors. In addition, we develop a parallel implementation of our model for large-scale applications. We demonstrate the effectiveness of our thLDA model on both a microblogging dataset and standard text collections including AP and RCV1 datasets.
[ { "version": "v1", "created": "Thu, 24 Jan 2013 02:02:13 GMT" }, { "version": "v2", "created": "Sat, 26 Jan 2013 18:00:19 GMT" } ]
2013-01-29T00:00:00
[ [ "Kang", "Jeon-Hyung", "" ], [ "Ma", "Jun", "" ], [ "Liu", "Yan", "" ] ]
TITLE: Transfer Topic Modeling with Ease and Scalability ABSTRACT: The increasing volume of short texts generated on social media sites, such as Twitter or Facebook, creates a great demand for effective and efficient topic modeling approaches. While latent Dirichlet allocation (LDA) can be applied, it is not optimal due to its weakness in handling short texts with fast-changing topics and scalability concerns. In this paper, we propose a transfer learning approach that utilizes abundant labeled documents from other domains (such as Yahoo! News or Wikipedia) to improve topic modeling, with better model fitting and result interpretation. Specifically, we develop Transfer Hierarchical LDA (thLDA) model, which incorporates the label information from other domains via informative priors. In addition, we develop a parallel implementation of our model for large-scale applications. We demonstrate the effectiveness of our thLDA model on both a microblogging dataset and standard text collections including AP and RCV1 datasets.
no_new_dataset
0.711982
1301.6553
Thomas Couronne
Thomas Couronne, Zbigniew Smoreda, Ana-Maria Olteanu
Chatty Mobiles:Individual mobility and communication patterns
NetMob 2011, Boston
null
null
null
cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Human mobility analysis is an important issue in social sciences, and mobility data are among the most sought-after sources of information in ur- Data ban studies, geography, transportation and territory management. In network sciences mobility studies have become popular in the past few years, especially using mobile phone location data. For preserving the customer privacy, datasets furnished by telecom operators are anonymized. At the same time, the large size of datasets often makes the task of calculating all observed trajectories very difficult and time-consuming. One solution could be to sample users. However, the fact of not having information about the mobile user makes the sampling delicate. Some researchers select randomly a sample of users from their dataset. Others try to optimize this method, for example, taking into account only users with a certain number or frequency of locations recorded. At the first glance, the second choice seems to be more efficient: having more individual traces makes the analysis more precise. However, the most frequently used CDR data (Call Detail Records) have location generated only at the moment of communication (call, SMS, data connection). Due to this fact, users mobility patterns cannot be precisely built upon their communication patterns. Hence, these data have evident short-comings both in terms of spatial and temporal scale. In this paper we propose to estimate the correlation between the users communication and mo- bility in order to better assess the bias of frequency based sampling. Using technical GSM network data (including communication but also independent mobility records), we will analyze the relationship between communication and mobility patterns.
[ { "version": "v1", "created": "Mon, 28 Jan 2013 14:19:48 GMT" } ]
2013-01-29T00:00:00
[ [ "Couronne", "Thomas", "" ], [ "Smoreda", "Zbigniew", "" ], [ "Olteanu", "Ana-Maria", "" ] ]
TITLE: Chatty Mobiles:Individual mobility and communication patterns ABSTRACT: Human mobility analysis is an important issue in social sciences, and mobility data are among the most sought-after sources of information in ur- Data ban studies, geography, transportation and territory management. In network sciences mobility studies have become popular in the past few years, especially using mobile phone location data. For preserving the customer privacy, datasets furnished by telecom operators are anonymized. At the same time, the large size of datasets often makes the task of calculating all observed trajectories very difficult and time-consuming. One solution could be to sample users. However, the fact of not having information about the mobile user makes the sampling delicate. Some researchers select randomly a sample of users from their dataset. Others try to optimize this method, for example, taking into account only users with a certain number or frequency of locations recorded. At the first glance, the second choice seems to be more efficient: having more individual traces makes the analysis more precise. However, the most frequently used CDR data (Call Detail Records) have location generated only at the moment of communication (call, SMS, data connection). Due to this fact, users mobility patterns cannot be precisely built upon their communication patterns. Hence, these data have evident short-comings both in terms of spatial and temporal scale. In this paper we propose to estimate the correlation between the users communication and mo- bility in order to better assess the bias of frequency based sampling. Using technical GSM network data (including communication but also independent mobility records), we will analyze the relationship between communication and mobility patterns.
no_new_dataset
0.706608
1301.3530
Charles Cadieu
Charles F. Cadieu, Ha Hong, Dan Yamins, Nicolas Pinto, Najib J. Majaj, James J. DiCarlo
The Neural Representation Benchmark and its Evaluation on Brain and Machine
The v1 version contained incorrectly computed kernel analysis curves and KA-AUC values for V4, IT, and the HT-L3 models. They have been corrected in this version
null
null
null
cs.NE cs.CV cs.LG q-bio.NC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A key requirement for the development of effective learning representations is their evaluation and comparison to representations we know to be effective. In natural sensory domains, the community has viewed the brain as a source of inspiration and as an implicit benchmark for success. However, it has not been possible to directly test representational learning algorithms directly against the representations contained in neural systems. Here, we propose a new benchmark for visual representations on which we have directly tested the neural representation in multiple visual cortical areas in macaque (utilizing data from [Majaj et al., 2012]), and on which any computer vision algorithm that produces a feature space can be tested. The benchmark measures the effectiveness of the neural or machine representation by computing the classification loss on the ordered eigendecomposition of a kernel matrix [Montavon et al., 2011]. In our analysis we find that the neural representation in visual area IT is superior to visual area V4. In our analysis of representational learning algorithms, we find that three-layer models approach the representational performance of V4 and the algorithm in [Le et al., 2012] surpasses the performance of V4. Impressively, we find that a recent supervised algorithm [Krizhevsky et al., 2012] achieves performance comparable to that of IT for an intermediate level of image variation difficulty, and surpasses IT at a higher difficulty level. We believe this result represents a major milestone: it is the first learning algorithm we have found that exceeds our current estimate of IT representation performance. We hope that this benchmark will assist the community in matching the representational performance of visual cortex and will serve as an initial rallying point for further correspondence between representations derived in brains and machines.
[ { "version": "v1", "created": "Tue, 15 Jan 2013 23:42:21 GMT" }, { "version": "v2", "created": "Fri, 25 Jan 2013 20:39:46 GMT" } ]
2013-01-28T00:00:00
[ [ "Cadieu", "Charles F.", "" ], [ "Hong", "Ha", "" ], [ "Yamins", "Dan", "" ], [ "Pinto", "Nicolas", "" ], [ "Majaj", "Najib J.", "" ], [ "DiCarlo", "James J.", "" ] ]
TITLE: The Neural Representation Benchmark and its Evaluation on Brain and Machine ABSTRACT: A key requirement for the development of effective learning representations is their evaluation and comparison to representations we know to be effective. In natural sensory domains, the community has viewed the brain as a source of inspiration and as an implicit benchmark for success. However, it has not been possible to directly test representational learning algorithms directly against the representations contained in neural systems. Here, we propose a new benchmark for visual representations on which we have directly tested the neural representation in multiple visual cortical areas in macaque (utilizing data from [Majaj et al., 2012]), and on which any computer vision algorithm that produces a feature space can be tested. The benchmark measures the effectiveness of the neural or machine representation by computing the classification loss on the ordered eigendecomposition of a kernel matrix [Montavon et al., 2011]. In our analysis we find that the neural representation in visual area IT is superior to visual area V4. In our analysis of representational learning algorithms, we find that three-layer models approach the representational performance of V4 and the algorithm in [Le et al., 2012] surpasses the performance of V4. Impressively, we find that a recent supervised algorithm [Krizhevsky et al., 2012] achieves performance comparable to that of IT for an intermediate level of image variation difficulty, and surpasses IT at a higher difficulty level. We believe this result represents a major milestone: it is the first learning algorithm we have found that exceeds our current estimate of IT representation performance. We hope that this benchmark will assist the community in matching the representational performance of visual cortex and will serve as an initial rallying point for further correspondence between representations derived in brains and machines.
no_new_dataset
0.705734
1301.5943
Lu\'is Filipe Te\'ofilo
Lu\'is Filipe Te\'ofilo, Luis Paulo Reis
Identifying Player\'s Strategies in No Limit Texas Hold\'em Poker through the Analysis of Individual Moves
null
null
null
null
cs.AI cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The development of competitive artificial Poker playing agents has proven to be a challenge, because agents must deal with unreliable information and deception which make it essential to model the opponents in order to achieve good results. This paper presents a methodology to develop opponent modeling techniques for Poker agents. The approach is based on applying clustering algorithms to a Poker game database in order to identify player types based on their actions. First, common game moves were identified by clustering all players\' moves. Then, player types were defined by calculating the frequency with which the players perform each type of movement. With the given dataset, 7 different types of players were identified with each one having at least one tactic that characterizes him. The identification of player types may improve the overall performance of Poker agents, because it helps the agents to predict the opponent\'s moves, by associating each opponent to a distinct cluster.
[ { "version": "v1", "created": "Fri, 25 Jan 2013 01:49:15 GMT" } ]
2013-01-28T00:00:00
[ [ "Teófilo", "Luís Filipe", "" ], [ "Reis", "Luis Paulo", "" ] ]
TITLE: Identifying Player\'s Strategies in No Limit Texas Hold\'em Poker through the Analysis of Individual Moves ABSTRACT: The development of competitive artificial Poker playing agents has proven to be a challenge, because agents must deal with unreliable information and deception which make it essential to model the opponents in order to achieve good results. This paper presents a methodology to develop opponent modeling techniques for Poker agents. The approach is based on applying clustering algorithms to a Poker game database in order to identify player types based on their actions. First, common game moves were identified by clustering all players\' moves. Then, player types were defined by calculating the frequency with which the players perform each type of movement. With the given dataset, 7 different types of players were identified with each one having at least one tactic that characterizes him. The identification of player types may improve the overall performance of Poker agents, because it helps the agents to predict the opponent\'s moves, by associating each opponent to a distinct cluster.
no_new_dataset
0.71081
1212.2142
Arnab Chatterjee
Arnab Chatterjee, Marija Mitrovi\'c and Santo Fortunato
Universality in voting behavior: an empirical analysis
19 pages, 10 figures, 8 tables. The elections data-sets can be downloaded from http://becs.aalto.fi/en/research/complex_systems/elections/
Scientific Reports 3, 1049 (2013)
null
null
physics.soc-ph cs.SI physics.data-an
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Election data represent a precious source of information to study human behavior at a large scale. In proportional elections with open lists, the number of votes received by a candidate, rescaled by the average performance of all competitors in the same party list, has the same distribution regardless of the country and the year of the election. Here we provide the first thorough assessment of this claim. We analyzed election datasets of 15 countries with proportional systems. We confirm that a class of nations with similar election rules fulfill the universality claim. Discrepancies from this trend in other countries with open-lists elections are always associated with peculiar differences in the election rules, which matter more than differences between countries and historical periods. Our analysis shows that the role of parties in the electoral performance of candidates is crucial: alternative scalings not taking into account party affiliations lead to poor results.
[ { "version": "v1", "created": "Mon, 10 Dec 2012 17:26:06 GMT" }, { "version": "v2", "created": "Thu, 24 Jan 2013 12:41:20 GMT" } ]
2013-01-25T00:00:00
[ [ "Chatterjee", "Arnab", "" ], [ "Mitrović", "Marija", "" ], [ "Fortunato", "Santo", "" ] ]
TITLE: Universality in voting behavior: an empirical analysis ABSTRACT: Election data represent a precious source of information to study human behavior at a large scale. In proportional elections with open lists, the number of votes received by a candidate, rescaled by the average performance of all competitors in the same party list, has the same distribution regardless of the country and the year of the election. Here we provide the first thorough assessment of this claim. We analyzed election datasets of 15 countries with proportional systems. We confirm that a class of nations with similar election rules fulfill the universality claim. Discrepancies from this trend in other countries with open-lists elections are always associated with peculiar differences in the election rules, which matter more than differences between countries and historical periods. Our analysis shows that the role of parties in the electoral performance of candidates is crucial: alternative scalings not taking into account party affiliations lead to poor results.
no_new_dataset
0.704402
1301.5399
Hoda Sadat Ayatollahi Tabatabaii
Hoda S. Ayatollahi Tabatabaii, Hamid R. Rabiee, Mohammad Hossein Rohban, Mostafa Salehi
Incorporating Betweenness Centrality in Compressive Sensing for Congestion Detection
null
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents a new Compressive Sensing (CS) scheme for detecting network congested links. We focus on decreasing the required number of measurements to detect all congested links in the context of network tomography. We have expanded the LASSO objective function by adding a new term corresponding to the prior knowledge based on the relationship between the congested links and the corresponding link Betweenness Centrality (BC). The accuracy of the proposed model is verified by simulations on two real datasets. The results demonstrate that our model outperformed the state-of-the-art CS based method with significant improvements in terms of F-Score.
[ { "version": "v1", "created": "Wed, 23 Jan 2013 04:12:08 GMT" } ]
2013-01-24T00:00:00
[ [ "Tabatabaii", "Hoda S. Ayatollahi", "" ], [ "Rabiee", "Hamid R.", "" ], [ "Rohban", "Mohammad Hossein", "" ], [ "Salehi", "Mostafa", "" ] ]
TITLE: Incorporating Betweenness Centrality in Compressive Sensing for Congestion Detection ABSTRACT: This paper presents a new Compressive Sensing (CS) scheme for detecting network congested links. We focus on decreasing the required number of measurements to detect all congested links in the context of network tomography. We have expanded the LASSO objective function by adding a new term corresponding to the prior knowledge based on the relationship between the congested links and the corresponding link Betweenness Centrality (BC). The accuracy of the proposed model is verified by simulations on two real datasets. The results demonstrate that our model outperformed the state-of-the-art CS based method with significant improvements in terms of F-Score.
no_new_dataset
0.71123
1204.4491
Huy Nguyen
Huy Nguyen, Rong Zheng
On Budgeted Influence Maximization in Social Networks
Submitted to JSAC NS
null
null
null
cs.SI physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Given a budget and arbitrary cost for selecting each node, the budgeted influence maximization (BIM) problem concerns selecting a set of seed nodes to disseminate some information that maximizes the total number of nodes influenced (termed as influence spread) in social networks at a total cost no more than the budget. Our proposed seed selection algorithm for the BIM problem guarantees an approximation ratio of (1 - 1/sqrt(e)). The seed selection algorithm needs to calculate the influence spread of candidate seed sets, which is known to be #P-complex. Identifying the linkage between the computation of marginal probabilities in Bayesian networks and the influence spread, we devise efficient heuristic algorithms for the latter problem. Experiments using both large-scale social networks and synthetically generated networks demonstrate superior performance of the proposed algorithm with moderate computation costs. Moreover, synthetic datasets allow us to vary the network parameters and gain important insights on the impact of graph structures on the performance of different algorithms.
[ { "version": "v1", "created": "Thu, 19 Apr 2012 22:50:48 GMT" }, { "version": "v2", "created": "Thu, 2 Aug 2012 05:02:19 GMT" }, { "version": "v3", "created": "Tue, 22 Jan 2013 07:01:49 GMT" } ]
2013-01-23T00:00:00
[ [ "Nguyen", "Huy", "" ], [ "Zheng", "Rong", "" ] ]
TITLE: On Budgeted Influence Maximization in Social Networks ABSTRACT: Given a budget and arbitrary cost for selecting each node, the budgeted influence maximization (BIM) problem concerns selecting a set of seed nodes to disseminate some information that maximizes the total number of nodes influenced (termed as influence spread) in social networks at a total cost no more than the budget. Our proposed seed selection algorithm for the BIM problem guarantees an approximation ratio of (1 - 1/sqrt(e)). The seed selection algorithm needs to calculate the influence spread of candidate seed sets, which is known to be #P-complex. Identifying the linkage between the computation of marginal probabilities in Bayesian networks and the influence spread, we devise efficient heuristic algorithms for the latter problem. Experiments using both large-scale social networks and synthetically generated networks demonstrate superior performance of the proposed algorithm with moderate computation costs. Moreover, synthetic datasets allow us to vary the network parameters and gain important insights on the impact of graph structures on the performance of different algorithms.
no_new_dataset
0.709837
1301.5088
Ian Goodfellow
Ian J. Goodfellow
Piecewise Linear Multilayer Perceptrons and Dropout
null
null
null
null
stat.ML cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose a new type of hidden layer for a multilayer perceptron, and demonstrate that it obtains the best reported performance for an MLP on the MNIST dataset.
[ { "version": "v1", "created": "Tue, 22 Jan 2013 07:10:34 GMT" } ]
2013-01-23T00:00:00
[ [ "Goodfellow", "Ian J.", "" ] ]
TITLE: Piecewise Linear Multilayer Perceptrons and Dropout ABSTRACT: We propose a new type of hidden layer for a multilayer perceptron, and demonstrate that it obtains the best reported performance for an MLP on the MNIST dataset.
no_new_dataset
0.713469
1301.5121
Alex Averbuch
Alex Averbuch, Martin Neumann
Partitioning Graph Databases - A Quantitative Evaluation
null
null
null
null
cs.DB cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Electronic data is growing at increasing rates, in both size and connectivity: the increasing presence of, and interest in, relationships between data. An example is the Twitter social network graph. Due to this growth demand is increasing for technologies that can process such data. Currently relational databases are the predominant technology, but they are poorly suited to processing connected data as they are optimized for index-intensive operations. Conversely, graph databases are optimized for graph computation. They link records by direct references, avoiding index lookups, and enabling retrieval of adjacent elements in constant time, regardless of graph size. However, as data volume increases these databases outgrow the resources of one computer and data partitioning becomes necessary. We evaluate the viability of using graph partitioning algorithms to partition graph databases. A prototype partitioned database was developed. Three partitioning algorithms explored and one implemented. Three graph datasets were used: two real and one synthetically generated. These were partitioned in various ways and the impact on database performance measured. We defined one synthetic access pattern per dataset and executed each on the partitioned datasets. Evaluation took place in a simulation environment, ensuring repeatability and allowing measurement of metrics like network traffic and load balance. Results show that compared to random partitioning the partitioning algorithm reduced traffic by 40-90%. Executing the algorithm intermittently during usage maintained partition quality, while requiring only 1% the computation of initial partitioning. Strong correlations were found between theoretic quality metrics and generated network traffic under non-uniform access patterns.
[ { "version": "v1", "created": "Tue, 22 Jan 2013 09:48:34 GMT" } ]
2013-01-23T00:00:00
[ [ "Averbuch", "Alex", "" ], [ "Neumann", "Martin", "" ] ]
TITLE: Partitioning Graph Databases - A Quantitative Evaluation ABSTRACT: Electronic data is growing at increasing rates, in both size and connectivity: the increasing presence of, and interest in, relationships between data. An example is the Twitter social network graph. Due to this growth demand is increasing for technologies that can process such data. Currently relational databases are the predominant technology, but they are poorly suited to processing connected data as they are optimized for index-intensive operations. Conversely, graph databases are optimized for graph computation. They link records by direct references, avoiding index lookups, and enabling retrieval of adjacent elements in constant time, regardless of graph size. However, as data volume increases these databases outgrow the resources of one computer and data partitioning becomes necessary. We evaluate the viability of using graph partitioning algorithms to partition graph databases. A prototype partitioned database was developed. Three partitioning algorithms explored and one implemented. Three graph datasets were used: two real and one synthetically generated. These were partitioned in various ways and the impact on database performance measured. We defined one synthetic access pattern per dataset and executed each on the partitioned datasets. Evaluation took place in a simulation environment, ensuring repeatability and allowing measurement of metrics like network traffic and load balance. Results show that compared to random partitioning the partitioning algorithm reduced traffic by 40-90%. Executing the algorithm intermittently during usage maintained partition quality, while requiring only 1% the computation of initial partitioning. Strong correlations were found between theoretic quality metrics and generated network traffic under non-uniform access patterns.
no_new_dataset
0.711782
1301.3753
Leif Johnson
Leif Johnson and Craig Corcoran
Switched linear encoding with rectified linear autoencoders
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Several recent results in machine learning have established formal connections between autoencoders---artificial neural network models that attempt to reproduce their inputs---and other coding models like sparse coding and K-means. This paper explores in depth an autoencoder model that is constructed using rectified linear activations on its hidden units. Our analysis builds on recent results to further unify the world of sparse linear coding models. We provide an intuitive interpretation of the behavior of these coding models and demonstrate this intuition using small, artificial datasets with known distributions.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 17:04:10 GMT" }, { "version": "v2", "created": "Sat, 19 Jan 2013 19:38:36 GMT" } ]
2013-01-22T00:00:00
[ [ "Johnson", "Leif", "" ], [ "Corcoran", "Craig", "" ] ]
TITLE: Switched linear encoding with rectified linear autoencoders ABSTRACT: Several recent results in machine learning have established formal connections between autoencoders---artificial neural network models that attempt to reproduce their inputs---and other coding models like sparse coding and K-means. This paper explores in depth an autoencoder model that is constructed using rectified linear activations on its hidden units. Our analysis builds on recent results to further unify the world of sparse linear coding models. We provide an intuitive interpretation of the behavior of these coding models and demonstrate this intuition using small, artificial datasets with known distributions.
no_new_dataset
0.708975
1301.4973
Philipp Ruemmer
Philipp R\"ummer (Uppsala University), Hossein Hojjat (EPFL Lausanne), Viktor Kuncak (EPFL Lausanne)
Disjunctive Interpolants for Horn-Clause Verification (Extended Technical Report)
null
null
null
null
cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
One of the main challenges in software verification is efficient and precise compositional analysis of programs with procedures and loops. Interpolation methods remain one of the most promising techniques for such verification, and are closely related to solving Horn clause constraints. We introduce a new notion of interpolation, disjunctive interpolation, which solve a more general class of problems in one step compared to previous notions of interpolants, such as tree interpolants or inductive sequences of interpolants. We present algorithms and complexity for construction of disjunctive interpolants, as well as their use within an abstraction-refinement loop. We have implemented Horn clause verification algorithms that use disjunctive interpolants and evaluate them on benchmarks expressed as Horn clauses over the theory of integer linear arithmetic.
[ { "version": "v1", "created": "Mon, 21 Jan 2013 20:13:32 GMT" } ]
2013-01-22T00:00:00
[ [ "Rümmer", "Philipp", "", "Uppsala University" ], [ "Hojjat", "Hossein", "", "EPFL Lausanne" ], [ "Kuncak", "Viktor", "", "EPFL Lausanne" ] ]
TITLE: Disjunctive Interpolants for Horn-Clause Verification (Extended Technical Report) ABSTRACT: One of the main challenges in software verification is efficient and precise compositional analysis of programs with procedures and loops. Interpolation methods remain one of the most promising techniques for such verification, and are closely related to solving Horn clause constraints. We introduce a new notion of interpolation, disjunctive interpolation, which solve a more general class of problems in one step compared to previous notions of interpolants, such as tree interpolants or inductive sequences of interpolants. We present algorithms and complexity for construction of disjunctive interpolants, as well as their use within an abstraction-refinement loop. We have implemented Horn clause verification algorithms that use disjunctive interpolants and evaluate them on benchmarks expressed as Horn clauses over the theory of integer linear arithmetic.
no_new_dataset
0.70724
1301.3127
Frederic Herbreteau
Fr\'ed\'eric Herbreteau (LaBRI), B. Srivathsan (RWTH), Igor Walukiewicz (LaBRI)
Lazy abstractions for timed automata
null
null
null
null
cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the reachability problem for timed automata. A standard solution to this problem involves computing a search tree whose nodes are abstractions of zones. For efficiency reasons, they are parametrized by the maximal lower and upper bounds (LU-bounds) occurring in the guards of the automaton. We propose an algorithm that is updating LU-bounds during exploration of the search tree. In order to keep them as small as possible, the bounds are refined only when they enable a transition that is impossible in the unabstracted system. So our algorithm can be seen as a kind of lazy CEGAR algorithm for timed automata. We show that on several standard benchmarks, the algorithm is capable of keeping very small LU-bounds, and in consequence reduce the search space substantially.
[ { "version": "v1", "created": "Mon, 14 Jan 2013 20:56:20 GMT" }, { "version": "v2", "created": "Fri, 18 Jan 2013 07:42:48 GMT" } ]
2013-01-21T00:00:00
[ [ "Herbreteau", "Frédéric", "", "LaBRI" ], [ "Srivathsan", "B.", "", "RWTH" ], [ "Walukiewicz", "Igor", "", "LaBRI" ] ]
TITLE: Lazy abstractions for timed automata ABSTRACT: We consider the reachability problem for timed automata. A standard solution to this problem involves computing a search tree whose nodes are abstractions of zones. For efficiency reasons, they are parametrized by the maximal lower and upper bounds (LU-bounds) occurring in the guards of the automaton. We propose an algorithm that is updating LU-bounds during exploration of the search tree. In order to keep them as small as possible, the bounds are refined only when they enable a transition that is impossible in the unabstracted system. So our algorithm can be seen as a kind of lazy CEGAR algorithm for timed automata. We show that on several standard benchmarks, the algorithm is capable of keeping very small LU-bounds, and in consequence reduce the search space substantially.
no_new_dataset
0.708421
1301.4490
Bharath Ramesh
Bharath Ramesh and Calvin J. Ribbens and Srinidhi Varadarajan
Regional Consistency: Programmability and Performance for Non-Cache-Coherent Systems
8 pages, 7 figures, 1 table; as submitted to CCGRID 2013
null
null
null
cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Parallel programmers face the often irreconcilable goals of programmability and performance. HPC systems use distributed memory for scalability, thereby sacrificing the programmability advantages of shared memory programming models. Furthermore, the rapid adoption of heterogeneous architectures, often with non-cache-coherent memory systems, has further increased the challenge of supporting shared memory programming models. Our primary objective is to define a memory consistency model that presents the familiar thread-based shared memory programming model, but allows good application performance on non-cache-coherent systems, including distributed memory clusters and accelerator-based systems. We propose regional consistency (RegC), a new consistency model that achieves this objective. Results on up to 256 processors for representative benchmarks demonstrate the potential of RegC in the context of our prototype distributed shared memory system.
[ { "version": "v1", "created": "Fri, 18 Jan 2013 20:40:42 GMT" } ]
2013-01-21T00:00:00
[ [ "Ramesh", "Bharath", "" ], [ "Ribbens", "Calvin J.", "" ], [ "Varadarajan", "Srinidhi", "" ] ]
TITLE: Regional Consistency: Programmability and Performance for Non-Cache-Coherent Systems ABSTRACT: Parallel programmers face the often irreconcilable goals of programmability and performance. HPC systems use distributed memory for scalability, thereby sacrificing the programmability advantages of shared memory programming models. Furthermore, the rapid adoption of heterogeneous architectures, often with non-cache-coherent memory systems, has further increased the challenge of supporting shared memory programming models. Our primary objective is to define a memory consistency model that presents the familiar thread-based shared memory programming model, but allows good application performance on non-cache-coherent systems, including distributed memory clusters and accelerator-based systems. We propose regional consistency (RegC), a new consistency model that achieves this objective. Results on up to 256 processors for representative benchmarks demonstrate the potential of RegC in the context of our prototype distributed shared memory system.
no_new_dataset
0.709667
1301.3844
Gregory F. Cooper
Gregory F. Cooper
A Bayesian Method for Causal Modeling and Discovery Under Selection
Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)
null
null
UAI-P-2000-PG-98-106
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper describes a Bayesian method for learning causal networks using samples that were selected in a non-random manner from a population of interest. Examples of data obtained by non-random sampling include convenience samples and case-control data in which a fixed number of samples with and without some condition is collected; such data are not uncommon. The paper describes a method for combining data under selection with prior beliefs in order to derive a posterior probability for a model of the causal processes that are generating the data in the population of interest. The priors include beliefs about the nature of the non-random sampling procedure. Although exact application of the method would be computationally intractable for most realistic datasets, efficient special-case and approximation methods are discussed. Finally, the paper describes how to combine learning under selection with previous methods for learning from observational and experimental data that are obtained on random samples of the population of interest. The net result is a Bayesian methodology that supports causal modeling and discovery from a rich mixture of different types of data.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:49:26 GMT" } ]
2013-01-18T00:00:00
[ [ "Cooper", "Gregory F.", "" ] ]
TITLE: A Bayesian Method for Causal Modeling and Discovery Under Selection ABSTRACT: This paper describes a Bayesian method for learning causal networks using samples that were selected in a non-random manner from a population of interest. Examples of data obtained by non-random sampling include convenience samples and case-control data in which a fixed number of samples with and without some condition is collected; such data are not uncommon. The paper describes a method for combining data under selection with prior beliefs in order to derive a posterior probability for a model of the causal processes that are generating the data in the population of interest. The priors include beliefs about the nature of the non-random sampling procedure. Although exact application of the method would be computationally intractable for most realistic datasets, efficient special-case and approximation methods are discussed. Finally, the paper describes how to combine learning under selection with previous methods for learning from observational and experimental data that are obtained on random samples of the population of interest. The net result is a Bayesian methodology that supports causal modeling and discovery from a rich mixture of different types of data.
no_new_dataset
0.710434
1301.3856
Nir Friedman
Nir Friedman, Daphne Koller
Being Bayesian about Network Structure
Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)
null
null
UAI-P-2000-PG-201-210
cs.LG cs.AI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In many domains, we are interested in analyzing the structure of the underlying distribution, e.g., whether one variable is a direct parent of the other. Bayesian model-selection attempts to find the MAP model and use its structure to answer these questions. However, when the amount of available data is modest, there might be many models that have non-negligible posterior. Thus, we want compute the Bayesian posterior of a feature, i.e., the total posterior probability of all models that contain it. In this paper, we propose a new approach for this task. We first show how to efficiently compute a sum over the exponential number of networks that are consistent with a fixed ordering over network variables. This allows us to compute, for a given ordering, both the marginal probability of the data and the posterior of a feature. We then use this result as the basis for an algorithm that approximates the Bayesian posterior of a feature. Our approach uses a Markov Chain Monte Carlo (MCMC) method, but over orderings rather than over network structures. The space of orderings is much smaller and more regular than the space of structures, and has a smoother posterior `landscape'. We present empirical results on synthetic and real-life datasets that compare our approach to full model averaging (when possible), to MCMC over network structures, and to a non-Bayesian bootstrap approach.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:50:14 GMT" } ]
2013-01-18T00:00:00
[ [ "Friedman", "Nir", "" ], [ "Koller", "Daphne", "" ] ]
TITLE: Being Bayesian about Network Structure ABSTRACT: In many domains, we are interested in analyzing the structure of the underlying distribution, e.g., whether one variable is a direct parent of the other. Bayesian model-selection attempts to find the MAP model and use its structure to answer these questions. However, when the amount of available data is modest, there might be many models that have non-negligible posterior. Thus, we want compute the Bayesian posterior of a feature, i.e., the total posterior probability of all models that contain it. In this paper, we propose a new approach for this task. We first show how to efficiently compute a sum over the exponential number of networks that are consistent with a fixed ordering over network variables. This allows us to compute, for a given ordering, both the marginal probability of the data and the posterior of a feature. We then use this result as the basis for an algorithm that approximates the Bayesian posterior of a feature. Our approach uses a Markov Chain Monte Carlo (MCMC) method, but over orderings rather than over network structures. The space of orderings is much smaller and more regular than the space of structures, and has a smoother posterior `landscape'. We present empirical results on synthetic and real-life datasets that compare our approach to full model averaging (when possible), to MCMC over network structures, and to a non-Bayesian bootstrap approach.
no_new_dataset
0.709837
1301.3884
Dmitry Y. Pavlov
Dmitry Y. Pavlov, Heikki Mannila, Padhraic Smyth
Probabilistic Models for Query Approximation with Large Sparse Binary Datasets
Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)
null
null
UAI-P-2000-PG-465-472
cs.AI cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Large sparse sets of binary transaction data with millions of records and thousands of attributes occur in various domains: customers purchasing products, users visiting web pages, and documents containing words are just three typical examples. Real-time query selectivity estimation (the problem of estimating the number of rows in the data satisfying a given predicate) is an important practical problem for such databases. We investigate the application of probabilistic models to this problem. In particular, we study a Markov random field (MRF) approach based on frequent sets and maximum entropy, and compare it to the independence model and the Chow-Liu tree model. We find that the MRF model provides substantially more accurate probability estimates than the other methods but is more expensive from a computational and memory viewpoint. To alleviate the computational requirements we show how one can apply bucket elimination and clique tree approaches to take advantage of structure in the models and in the queries. We provide experimental results on two large real-world transaction datasets.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:52:06 GMT" } ]
2013-01-18T00:00:00
[ [ "Pavlov", "Dmitry Y.", "" ], [ "Mannila", "Heikki", "" ], [ "Smyth", "Padhraic", "" ] ]
TITLE: Probabilistic Models for Query Approximation with Large Sparse Binary Datasets ABSTRACT: Large sparse sets of binary transaction data with millions of records and thousands of attributes occur in various domains: customers purchasing products, users visiting web pages, and documents containing words are just three typical examples. Real-time query selectivity estimation (the problem of estimating the number of rows in the data satisfying a given predicate) is an important practical problem for such databases. We investigate the application of probabilistic models to this problem. In particular, we study a Markov random field (MRF) approach based on frequent sets and maximum entropy, and compare it to the independence model and the Chow-Liu tree model. We find that the MRF model provides substantially more accurate probability estimates than the other methods but is more expensive from a computational and memory viewpoint. To alleviate the computational requirements we show how one can apply bucket elimination and clique tree approaches to take advantage of structure in the models and in the queries. We provide experimental results on two large real-world transaction datasets.
no_new_dataset
0.70939
1301.3891
Marc Sebban
Marc Sebban, Richard Nock
Combining Feature and Prototype Pruning by Uncertainty Minimization
Appears in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence (UAI2000)
null
null
UAI-P-2000-PG-533-540
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We focus in this paper on dataset reduction techniques for use in k-nearest neighbor classification. In such a context, feature and prototype selections have always been independently treated by the standard storage reduction algorithms. While this certifying is theoretically justified by the fact that each subproblem is NP-hard, we assume in this paper that a joint storage reduction is in fact more intuitive and can in practice provide better results than two independent processes. Moreover, it avoids a lot of distance calculations by progressively removing useless instances during the feature pruning. While standard selection algorithms often optimize the accuracy to discriminate the set of solutions, we use in this paper a criterion based on an uncertainty measure within a nearest-neighbor graph. This choice comes from recent results that have proven that accuracy is not always the suitable criterion to optimize. In our approach, a feature or an instance is removed if its deletion improves information of the graph. Numerous experiments are presented in this paper and a statistical analysis shows the relevance of our approach, and its tolerance in the presence of noise.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 15:52:33 GMT" } ]
2013-01-18T00:00:00
[ [ "Sebban", "Marc", "" ], [ "Nock", "Richard", "" ] ]
TITLE: Combining Feature and Prototype Pruning by Uncertainty Minimization ABSTRACT: We focus in this paper on dataset reduction techniques for use in k-nearest neighbor classification. In such a context, feature and prototype selections have always been independently treated by the standard storage reduction algorithms. While this certifying is theoretically justified by the fact that each subproblem is NP-hard, we assume in this paper that a joint storage reduction is in fact more intuitive and can in practice provide better results than two independent processes. Moreover, it avoids a lot of distance calculations by progressively removing useless instances during the feature pruning. While standard selection algorithms often optimize the accuracy to discriminate the set of solutions, we use in this paper a criterion based on an uncertainty measure within a nearest-neighbor graph. This choice comes from recent results that have proven that accuracy is not always the suitable criterion to optimize. In our approach, a feature or an instance is removed if its deletion improves information of the graph. Numerous experiments are presented in this paper and a statistical analysis shows the relevance of our approach, and its tolerance in the presence of noise.
no_new_dataset
0.711437
1301.3991
Zhenghong Chen
Zhenghong Chen, Xiaoxian Tang, Bican Xia
Generic Regular Decompositions for Parametric Polynomial Systems
It is the latest version. arXiv admin note: text overlap with arXiv:1208.6112
null
null
null
cs.SC cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents a generalization of our earlier work in [19]. In this paper, the two concepts, generic regular decomposition (GRD) and regular-decomposition-unstable (RDU) variety introduced in [19] for generic zero-dimensional systems, are extended to the case where the parametric systems are not necessarily zero-dimensional. An algorithm is provided to compute GRDs and the associated RDU varieties of parametric systems simultaneously on the basis of the algorithm for generic zero-dimensional systems proposed in [19]. Then the solutions of any parametric system can be represented by the solutions of finitely many regular systems and the decomposition is stable at any parameter value in the complement of the associated RDU variety of the parameter space. The related definitions and the results presented in [19] are also generalized and a further discussion on RDU varieties is given from an experimental point of view. The new algorithm has been implemented on the basis of DISCOVERER with Maple 16 and experimented with a number of benchmarks from the literature.
[ { "version": "v1", "created": "Thu, 17 Jan 2013 06:16:02 GMT" } ]
2013-01-18T00:00:00
[ [ "Chen", "Zhenghong", "" ], [ "Tang", "Xiaoxian", "" ], [ "Xia", "Bican", "" ] ]
TITLE: Generic Regular Decompositions for Parametric Polynomial Systems ABSTRACT: This paper presents a generalization of our earlier work in [19]. In this paper, the two concepts, generic regular decomposition (GRD) and regular-decomposition-unstable (RDU) variety introduced in [19] for generic zero-dimensional systems, are extended to the case where the parametric systems are not necessarily zero-dimensional. An algorithm is provided to compute GRDs and the associated RDU varieties of parametric systems simultaneously on the basis of the algorithm for generic zero-dimensional systems proposed in [19]. Then the solutions of any parametric system can be represented by the solutions of finitely many regular systems and the decomposition is stable at any parameter value in the complement of the associated RDU variety of the parameter space. The related definitions and the results presented in [19] are also generalized and a further discussion on RDU varieties is given from an experimental point of view. The new algorithm has been implemented on the basis of DISCOVERER with Maple 16 and experimented with a number of benchmarks from the literature.
no_new_dataset
0.709416
1301.4028
Michael Schreiber
Michael Schreiber
Do we need the g-index?
7 pages, 3 figures accepted for publication in Journal of the American Society for Information Science and Technology
null
null
null
physics.soc-ph cs.DL
http://creativecommons.org/licenses/by-nc-sa/3.0/
Using a very small sample of 8 datasets it was recently shown by De Visscher (2011) that the g-index is very close to the square root of the total number of citations. It was argued that there is no bibliometrically meaningful difference. Using another somewhat larger empirical sample of 26 datasets I show that the difference may be larger and I argue in favor of the g-index.
[ { "version": "v1", "created": "Thu, 17 Jan 2013 09:45:27 GMT" } ]
2013-01-18T00:00:00
[ [ "Schreiber", "Michael", "" ] ]
TITLE: Do we need the g-index? ABSTRACT: Using a very small sample of 8 datasets it was recently shown by De Visscher (2011) that the g-index is very close to the square root of the total number of citations. It was argued that there is no bibliometrically meaningful difference. Using another somewhat larger empirical sample of 26 datasets I show that the difference may be larger and I argue in favor of the g-index.
no_new_dataset
0.709466
1301.4171
Jason Weston
Jason Weston, Ron Weiss, Hector Yee
Affinity Weighted Embedding
null
null
null
null
cs.IR cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Supervised (linear) embedding models like Wsabie and PSI have proven successful at ranking, recommendation and annotation tasks. However, despite being scalable to large datasets they do not take full advantage of the extra data due to their linear nature, and typically underfit. We propose a new class of models which aim to provide improved performance while retaining many of the benefits of the existing class of embedding models. Our new approach works by iteratively learning a linear embedding model where the next iteration's features and labels are reweighted as a function of the previous iteration. We describe several variants of the family, and give some initial results.
[ { "version": "v1", "created": "Thu, 17 Jan 2013 17:46:27 GMT" } ]
2013-01-18T00:00:00
[ [ "Weston", "Jason", "" ], [ "Weiss", "Ron", "" ], [ "Yee", "Hector", "" ] ]
TITLE: Affinity Weighted Embedding ABSTRACT: Supervised (linear) embedding models like Wsabie and PSI have proven successful at ranking, recommendation and annotation tasks. However, despite being scalable to large datasets they do not take full advantage of the extra data due to their linear nature, and typically underfit. We propose a new class of models which aim to provide improved performance while retaining many of the benefits of the existing class of embedding models. Our new approach works by iteratively learning a linear embedding model where the next iteration's features and labels are reweighted as a function of the previous iteration. We describe several variants of the family, and give some initial results.
no_new_dataset
0.710245
1207.0166
Claudio Gentile
Claudio Gentile and Francesco Orabona
On Multilabel Classification and Ranking with Partial Feedback
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a novel multilabel/ranking algorithm working in partial information settings. The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation. We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models. We show O(T^{1/2} log T) regret bounds, which improve in several ways on the existing results. We test the effectiveness of our upper-confidence scheme by contrasting against full-information baselines on real-world multilabel datasets, often obtaining comparable performance.
[ { "version": "v1", "created": "Sat, 30 Jun 2012 23:07:03 GMT" }, { "version": "v2", "created": "Tue, 20 Nov 2012 16:48:22 GMT" }, { "version": "v3", "created": "Wed, 16 Jan 2013 19:19:34 GMT" } ]
2013-01-17T00:00:00
[ [ "Gentile", "Claudio", "" ], [ "Orabona", "Francesco", "" ] ]
TITLE: On Multilabel Classification and Ranking with Partial Feedback ABSTRACT: We present a novel multilabel/ranking algorithm working in partial information settings. The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation. We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models. We show O(T^{1/2} log T) regret bounds, which improve in several ways on the existing results. We test the effectiveness of our upper-confidence scheme by contrasting against full-information baselines on real-world multilabel datasets, often obtaining comparable performance.
no_new_dataset
0.709453
1208.6112
Xiaoxian Tang
Xiaoxian Tang, Zhenghong Chen, Bican Xia
Generic Regular Decompositions for Generic Zero-Dimensional Systems
Accepted by Science China: Information Science
null
null
null
cs.SC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Two new concepts, generic regular decomposition and regular-decomposition-unstable (RDU) variety for generic zero-dimensional systems, are introduced in this paper and an algorithm is proposed for computing a generic regular decomposition and the associated RDU variety of a given generic zero-dimensional system simultaneously. The solutions of the given system can be expressed by finitely many zero-dimensional regular chains if the parameter value is not on the RDU variety. The so called weakly relatively simplicial decomposition plays a crucial role in the algorithm, which is based on the theories of subresultant chains. Furthermore, the algorithm can be naturally adopted to compute a non-redundant Wu's decomposition and the decomposition is stable at any parameter value that is not on the RDU variety. The algorithm has been implemented with Maple 15 and experimented with a number of benchmarks from the literature. Empirical results are also presented to show the good performance of the algorithm.
[ { "version": "v1", "created": "Thu, 30 Aug 2012 08:43:38 GMT" }, { "version": "v2", "created": "Fri, 31 Aug 2012 08:27:48 GMT" }, { "version": "v3", "created": "Mon, 3 Sep 2012 06:27:13 GMT" }, { "version": "v4", "created": "Wed, 16 Jan 2013 03:16:31 GMT" } ]
2013-01-17T00:00:00
[ [ "Tang", "Xiaoxian", "" ], [ "Chen", "Zhenghong", "" ], [ "Xia", "Bican", "" ] ]
TITLE: Generic Regular Decompositions for Generic Zero-Dimensional Systems ABSTRACT: Two new concepts, generic regular decomposition and regular-decomposition-unstable (RDU) variety for generic zero-dimensional systems, are introduced in this paper and an algorithm is proposed for computing a generic regular decomposition and the associated RDU variety of a given generic zero-dimensional system simultaneously. The solutions of the given system can be expressed by finitely many zero-dimensional regular chains if the parameter value is not on the RDU variety. The so called weakly relatively simplicial decomposition plays a crucial role in the algorithm, which is based on the theories of subresultant chains. Furthermore, the algorithm can be naturally adopted to compute a non-redundant Wu's decomposition and the decomposition is stable at any parameter value that is not on the RDU variety. The algorithm has been implemented with Maple 15 and experimented with a number of benchmarks from the literature. Empirical results are also presented to show the good performance of the algorithm.
no_new_dataset
0.707234
1301.3385
Steven Young
Steven R. Young and Itamar Arel
Recurrent Online Clustering as a Spatio-Temporal Feature Extractor in DeSTIN
3 pages, 2 figures, Submitted to ICLR 2013
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents a basic enhancement to the DeSTIN deep learning architecture by replacing the explicitly calculated transition tables that are used to capture temporal features with a simpler, more scalable mechanism. This mechanism uses feedback of state information to cluster over a space comprised of both the spatial input and the current state. The resulting architecture achieves state-of-the-art results on the MNIST classification benchmark.
[ { "version": "v1", "created": "Tue, 15 Jan 2013 15:34:07 GMT" }, { "version": "v2", "created": "Wed, 16 Jan 2013 14:56:44 GMT" } ]
2013-01-17T00:00:00
[ [ "Young", "Steven R.", "" ], [ "Arel", "Itamar", "" ] ]
TITLE: Recurrent Online Clustering as a Spatio-Temporal Feature Extractor in DeSTIN ABSTRACT: This paper presents a basic enhancement to the DeSTIN deep learning architecture by replacing the explicitly calculated transition tables that are used to capture temporal features with a simpler, more scalable mechanism. This mechanism uses feedback of state information to cluster over a space comprised of both the spatial input and the current state. The resulting architecture achieves state-of-the-art results on the MNIST classification benchmark.
no_new_dataset
0.71039
1301.3524
Indre Zliobaite
Indre Zliobaite
How good is the Electricity benchmark for evaluating concept drift adaptation
2 pages of content, 1 appendix, 2 figures
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this correspondence, we will point out a problem with testing adaptive classifiers on autocorrelated data. In such a case random change alarms may boost the accuracy figures. Hence, we cannot be sure if the adaptation is working well.
[ { "version": "v1", "created": "Tue, 15 Jan 2013 22:51:40 GMT" } ]
2013-01-17T00:00:00
[ [ "Zliobaite", "Indre", "" ] ]
TITLE: How good is the Electricity benchmark for evaluating concept drift adaptation ABSTRACT: In this correspondence, we will point out a problem with testing adaptive classifiers on autocorrelated data. In such a case random change alarms may boost the accuracy figures. Hence, we cannot be sure if the adaptation is working well.
no_new_dataset
0.708452
1301.3528
Momiao Xiong
Momiao Xiong and Long Ma
An Efficient Sufficient Dimension Reduction Method for Identifying Genetic Variants of Clinical Significance
null
null
null
null
q-bio.GN cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Fast and cheaper next generation sequencing technologies will generate unprecedentedly massive and highly-dimensional genomic and epigenomic variation data. In the near future, a routine part of medical record will include the sequenced genomes. A fundamental question is how to efficiently extract genomic and epigenomic variants of clinical utility which will provide information for optimal wellness and interference strategies. Traditional paradigm for identifying variants of clinical validity is to test association of the variants. However, significantly associated genetic variants may or may not be usefulness for diagnosis and prognosis of diseases. Alternative to association studies for finding genetic variants of predictive utility is to systematically search variants that contain sufficient information for phenotype prediction. To achieve this, we introduce concepts of sufficient dimension reduction and coordinate hypothesis which project the original high dimensional data to very low dimensional space while preserving all information on response phenotypes. We then formulate clinically significant genetic variant discovery problem into sparse SDR problem and develop algorithms that can select significant genetic variants from up to or even ten millions of predictors with the aid of dividing SDR for whole genome into a number of subSDR problems defined for genomic regions. The sparse SDR is in turn formulated as sparse optimal scoring problem, but with penalty which can remove row vectors from the basis matrix. To speed up computation, we develop the modified alternating direction method for multipliers to solve the sparse optimal scoring problem which can easily be implemented in parallel. To illustrate its application, the proposed method is applied to simulation data and the NHLBI's Exome Sequencing Project dataset
[ { "version": "v1", "created": "Tue, 15 Jan 2013 23:19:14 GMT" } ]
2013-01-17T00:00:00
[ [ "Xiong", "Momiao", "" ], [ "Ma", "Long", "" ] ]
TITLE: An Efficient Sufficient Dimension Reduction Method for Identifying Genetic Variants of Clinical Significance ABSTRACT: Fast and cheaper next generation sequencing technologies will generate unprecedentedly massive and highly-dimensional genomic and epigenomic variation data. In the near future, a routine part of medical record will include the sequenced genomes. A fundamental question is how to efficiently extract genomic and epigenomic variants of clinical utility which will provide information for optimal wellness and interference strategies. Traditional paradigm for identifying variants of clinical validity is to test association of the variants. However, significantly associated genetic variants may or may not be usefulness for diagnosis and prognosis of diseases. Alternative to association studies for finding genetic variants of predictive utility is to systematically search variants that contain sufficient information for phenotype prediction. To achieve this, we introduce concepts of sufficient dimension reduction and coordinate hypothesis which project the original high dimensional data to very low dimensional space while preserving all information on response phenotypes. We then formulate clinically significant genetic variant discovery problem into sparse SDR problem and develop algorithms that can select significant genetic variants from up to or even ten millions of predictors with the aid of dividing SDR for whole genome into a number of subSDR problems defined for genomic regions. The sparse SDR is in turn formulated as sparse optimal scoring problem, but with penalty which can remove row vectors from the basis matrix. To speed up computation, we develop the modified alternating direction method for multipliers to solve the sparse optimal scoring problem which can easily be implemented in parallel. To illustrate its application, the proposed method is applied to simulation data and the NHLBI's Exome Sequencing Project dataset
no_new_dataset
0.708994
1301.3539
Yoonseop Kang
Yoonseop Kang and Seungjin Choi
Learning Features with Structure-Adapting Multi-view Exponential Family Harmoniums
3 pages, 2 figures, ICLR2013 workshop track submission
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We proposea graphical model for multi-view feature extraction that automatically adapts its structure to achieve better representation of data distribution. The proposed model, structure-adapting multi-view harmonium (SA-MVH) has switch parameters that control the connection between hidden nodes and input views, and learn the switch parameter while training. Numerical experiments on synthetic and a real-world dataset demonstrate the useful behavior of the SA-MVH, compared to existing multi-view feature extraction methods.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 01:07:38 GMT" } ]
2013-01-17T00:00:00
[ [ "Kang", "Yoonseop", "" ], [ "Choi", "Seungjin", "" ] ]
TITLE: Learning Features with Structure-Adapting Multi-view Exponential Family Harmoniums ABSTRACT: We proposea graphical model for multi-view feature extraction that automatically adapts its structure to achieve better representation of data distribution. The proposed model, structure-adapting multi-view harmonium (SA-MVH) has switch parameters that control the connection between hidden nodes and input views, and learn the switch parameter while training. Numerical experiments on synthetic and a real-world dataset demonstrate the useful behavior of the SA-MVH, compared to existing multi-view feature extraction methods.
no_new_dataset
0.711619
1301.3557
Matthew Zeiler
Matthew D. Zeiler and Rob Fergus
Stochastic Pooling for Regularization of Deep Convolutional Neural Networks
9 pages
null
null
null
cs.LG cs.NE stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce a simple and effective method for regularizing large convolutional neural networks. We replace the conventional deterministic pooling operations with a stochastic procedure, randomly picking the activation within each pooling region according to a multinomial distribution, given by the activities within the pooling region. The approach is hyper-parameter free and can be combined with other regularization approaches, such as dropout and data augmentation. We achieve state-of-the-art performance on four image datasets, relative to other approaches that do not utilize data augmentation.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 02:12:07 GMT" } ]
2013-01-17T00:00:00
[ [ "Zeiler", "Matthew D.", "" ], [ "Fergus", "Rob", "" ] ]
TITLE: Stochastic Pooling for Regularization of Deep Convolutional Neural Networks ABSTRACT: We introduce a simple and effective method for regularizing large convolutional neural networks. We replace the conventional deterministic pooling operations with a stochastic procedure, randomly picking the activation within each pooling region according to a multinomial distribution, given by the activities within the pooling region. The approach is hyper-parameter free and can be combined with other regularization approaches, such as dropout and data augmentation. We achieve state-of-the-art performance on four image datasets, relative to other approaches that do not utilize data augmentation.
no_new_dataset
0.710879
1301.3744
Tim Vines
Timothy H. Vines, Rose L. Andrew, Dan G. Bock, Michelle T. Franklin, Kimberly J. Gilbert, Nolan C. Kane, Jean-S\'ebastien Moore, Brook T. Moyers, S\'ebastien Renaut, Diana J. Rennison, Thor Veen, Sam Yeaman
Mandated data archiving greatly improves access to research data
null
null
10.1096/fj.12-218164
null
cs.DL physics.soc-ph q-bio.QM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The data underlying scientific papers should be accessible to researchers both now and in the future, but how best can we ensure that these data are available? Here we examine the effectiveness of four approaches to data archiving: no stated archiving policy, recommending (but not requiring) archiving, and two versions of mandating data deposition at acceptance. We control for differences between data types by trying to obtain data from papers that use a single, widespread population genetic analysis, STRUCTURE. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand-fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. We also assessed the effectiveness of asking for data directly from authors and obtained over half of the requested datasets, albeit with about 8 days delay and some disagreement with authors. Given the long term benefits of data accessibility to the academic community, we believe that journal based mandatory data archiving policies and mandatory data availability statements should be more widely adopted.
[ { "version": "v1", "created": "Wed, 16 Jan 2013 16:22:26 GMT" } ]
2013-01-17T00:00:00
[ [ "Vines", "Timothy H.", "" ], [ "Andrew", "Rose L.", "" ], [ "Bock", "Dan G.", "" ], [ "Franklin", "Michelle T.", "" ], [ "Gilbert", "Kimberly J.", "" ], [ "Kane", "Nolan C.", "" ], [ "Moore", "Jean-Sébastien", "" ], [ "Moyers", "Brook T.", "" ], [ "Renaut", "Sébastien", "" ], [ "Rennison", "Diana J.", "" ], [ "Veen", "Thor", "" ], [ "Yeaman", "Sam", "" ] ]
TITLE: Mandated data archiving greatly improves access to research data ABSTRACT: The data underlying scientific papers should be accessible to researchers both now and in the future, but how best can we ensure that these data are available? Here we examine the effectiveness of four approaches to data archiving: no stated archiving policy, recommending (but not requiring) archiving, and two versions of mandating data deposition at acceptance. We control for differences between data types by trying to obtain data from papers that use a single, widespread population genetic analysis, STRUCTURE. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand-fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. At one extreme, we found that mandated data archiving policies that require the inclusion of a data availability statement in the manuscript improve the odds of finding the data online almost a thousand fold compared to having no policy. However, archiving rates at journals with less stringent policies were only very slightly higher than those with no policy at all. We also assessed the effectiveness of asking for data directly from authors and obtained over half of the requested datasets, albeit with about 8 days delay and some disagreement with authors. Given the long term benefits of data accessibility to the academic community, we believe that journal based mandatory data archiving policies and mandatory data availability statements should be more widely adopted.
no_new_dataset
0.71262
1301.0239
Rodrigo Aldecoa
Rodrigo Aldecoa and Ignacio Mar\'in
Surprise maximization reveals the community structure of complex networks
8 pages, 6 figures, 2 extra tables. Accepted for publication in Scientific Reports
Scientific Reports 3, 1060 (2013)
10.1038/srep01060
null
cs.SI cond-mat.stat-mech physics.soc-ph q-bio.MN
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
How to determine the community structure of complex networks is an open question. It is critical to establish the best strategies for community detection in networks of unknown structure. Here, using standard synthetic benchmarks, we show that none of the algorithms hitherto developed for community structure characterization perform optimally. Significantly, evaluating the results according to their modularity, the most popular measure of the quality of a partition, systematically provides mistaken solutions. However, a novel quality function, called Surprise, can be used to elucidate which is the optimal division into communities. Consequently, we show that the best strategy to find the community structure of all the networks examined involves choosing among the solutions provided by multiple algorithms the one with the highest Surprise value. We conclude that Surprise maximization precisely reveals the community structure of complex networks.
[ { "version": "v1", "created": "Wed, 2 Jan 2013 14:24:03 GMT" }, { "version": "v2", "created": "Thu, 3 Jan 2013 13:55:12 GMT" } ]
2013-01-15T00:00:00
[ [ "Aldecoa", "Rodrigo", "" ], [ "Marín", "Ignacio", "" ] ]
TITLE: Surprise maximization reveals the community structure of complex networks ABSTRACT: How to determine the community structure of complex networks is an open question. It is critical to establish the best strategies for community detection in networks of unknown structure. Here, using standard synthetic benchmarks, we show that none of the algorithms hitherto developed for community structure characterization perform optimally. Significantly, evaluating the results according to their modularity, the most popular measure of the quality of a partition, systematically provides mistaken solutions. However, a novel quality function, called Surprise, can be used to elucidate which is the optimal division into communities. Consequently, we show that the best strategy to find the community structure of all the networks examined involves choosing among the solutions provided by multiple algorithms the one with the highest Surprise value. We conclude that Surprise maximization precisely reveals the community structure of complex networks.
no_new_dataset
0.710201
1301.2659
Fabrice Rossi
Romain Guigour\`es, Marc Boull\'e, Fabrice Rossi (SAMM)
A Triclustering Approach for Time Evolving Graphs
null
Co-clustering and Applications International Conference on Data Mining Workshop, Brussels : Belgium (2012)
10.1109/ICDMW.2012.61
null
cs.LG cs.SI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper introduces a novel technique to track structures in time evolving graphs. The method is based on a parameter free approach for three-dimensional co-clustering of the source vertices, the target vertices and the time. All these features are simultaneously segmented in order to build time segments and clusters of vertices whose edge distributions are similar and evolve in the same way over the time segments. The main novelty of this approach lies in that the time segments are directly inferred from the evolution of the edge distribution between the vertices, thus not requiring the user to make an a priori discretization. Experiments conducted on a synthetic dataset illustrate the good behaviour of the technique, and a study of a real-life dataset shows the potential of the proposed approach for exploratory data analysis.
[ { "version": "v1", "created": "Sat, 12 Jan 2013 07:51:14 GMT" } ]
2013-01-15T00:00:00
[ [ "Guigourès", "Romain", "", "SAMM" ], [ "Boullé", "Marc", "", "SAMM" ], [ "Rossi", "Fabrice", "", "SAMM" ] ]
TITLE: A Triclustering Approach for Time Evolving Graphs ABSTRACT: This paper introduces a novel technique to track structures in time evolving graphs. The method is based on a parameter free approach for three-dimensional co-clustering of the source vertices, the target vertices and the time. All these features are simultaneously segmented in order to build time segments and clusters of vertices whose edge distributions are similar and evolve in the same way over the time segments. The main novelty of this approach lies in that the time segments are directly inferred from the evolution of the edge distribution between the vertices, thus not requiring the user to make an a priori discretization. Experiments conducted on a synthetic dataset illustrate the good behaviour of the technique, and a study of a real-life dataset shows the potential of the proposed approach for exploratory data analysis.
no_new_dataset
0.710478
1301.2698
Jiankou Li
Jiankou Li
Evaluating community structure in large network with random walks
10 pages, 5 figures
null
null
null
cs.SI physics.soc-ph
http://creativecommons.org/licenses/by/3.0/
Community structure is one of the most important properties of networks. Most community algorithms are not suitable for large networks because of their time consuming. In fact there are lots of networks with millons even billons of nodes. In such case, most algorithms running in time O(n2logn) or even larger are not practical. What we need are linear or approximately linear time algorithm. Rising in response to such needs, we propose a quick methods to evaluate community structure in networks and then put forward a local community algorithm with nearly linear time based on random walks. Using our community evaluating measure, we could find some difference results from measures used before, i.e., the Newman Modularity. Our algorithm are effective in small benchmark networks with small less accuracy than more complex algorithms but a great of advantage in time consuming for large networks, especially super large networks.
[ { "version": "v1", "created": "Sat, 12 Jan 2013 16:49:08 GMT" } ]
2013-01-15T00:00:00
[ [ "Li", "Jiankou", "" ] ]
TITLE: Evaluating community structure in large network with random walks ABSTRACT: Community structure is one of the most important properties of networks. Most community algorithms are not suitable for large networks because of their time consuming. In fact there are lots of networks with millons even billons of nodes. In such case, most algorithms running in time O(n2logn) or even larger are not practical. What we need are linear or approximately linear time algorithm. Rising in response to such needs, we propose a quick methods to evaluate community structure in networks and then put forward a local community algorithm with nearly linear time based on random walks. Using our community evaluating measure, we could find some difference results from measures used before, i.e., the Newman Modularity. Our algorithm are effective in small benchmark networks with small less accuracy than more complex algorithms but a great of advantage in time consuming for large networks, especially super large networks.
no_new_dataset
0.710415
1301.2785
Rafi Muhammad
Muhammad Rafi, Mohammad Shahid Shaikh
A comparison of SVM and RVM for Document Classification
ICoCSIM 2012, Medan Indonesia
null
null
null
cs.IR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Document classification is a task of assigning a new unclassified document to one of the predefined set of classes. The content based document classification uses the content of the document with some weighting criteria to assign it to one of the predefined classes. It is a major task in library science, electronic document management systems and information sciences. This paper investigates document classification by using two different classification techniques (1) Support Vector Machine (SVM) and (2) Relevance Vector Machine (RVM). SVM is a supervised machine learning technique that can be used for classification task. In its basic form, SVM represents the instances of the data into space and tries to separate the distinct classes by a maximum possible wide gap (hyper plane) that separates the classes. On the other hand RVM uses probabilistic measure to define this separation space. RVM uses Bayesian inference to obtain succinct solution, thus RVM uses significantly fewer basis functions. Experimental studies on three standard text classification datasets reveal that although RVM takes more training time, its classification is much better as compared to SVM.
[ { "version": "v1", "created": "Sun, 13 Jan 2013 15:58:09 GMT" } ]
2013-01-15T00:00:00
[ [ "Rafi", "Muhammad", "" ], [ "Shaikh", "Mohammad Shahid", "" ] ]
TITLE: A comparison of SVM and RVM for Document Classification ABSTRACT: Document classification is a task of assigning a new unclassified document to one of the predefined set of classes. The content based document classification uses the content of the document with some weighting criteria to assign it to one of the predefined classes. It is a major task in library science, electronic document management systems and information sciences. This paper investigates document classification by using two different classification techniques (1) Support Vector Machine (SVM) and (2) Relevance Vector Machine (RVM). SVM is a supervised machine learning technique that can be used for classification task. In its basic form, SVM represents the instances of the data into space and tries to separate the distinct classes by a maximum possible wide gap (hyper plane) that separates the classes. On the other hand RVM uses probabilistic measure to define this separation space. RVM uses Bayesian inference to obtain succinct solution, thus RVM uses significantly fewer basis functions. Experimental studies on three standard text classification datasets reveal that although RVM takes more training time, its classification is much better as compared to SVM.
no_new_dataset
0.710051
1301.3120
Richard Darst
Richard K. Darst David R. Reichman Peter Ronhovde and Zohar Nussinov
An edge density definition of overlapping and weighted graph communities
22 pages, 12 figures
null
null
null
physics.soc-ph cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Community detection in networks refers to the process of seeking strongly internally connected groups of nodes which are weakly externally connected. In this work, we introduce and study a community definition based on internal edge density. Beginning with the simple concept that edge density equals number of edges divided by maximal number of edges, we apply this definition to a variety of node and community arrangements to show that our definition yields sensible results. Our community definition is equivalent to that of the Absolute Potts Model community detection method (Phys. Rev. E 81, 046114 (2010)), and the performance of that method validates the usefulness of our definition across a wide variety of network types. We discuss how this definition can be extended to weighted, and multigraphs, and how the definition is capable of handling overlapping communities and local algorithms. We further validate our definition against the recently proposed Affiliation Graph Model (arXiv:1205.6228 [cs.SI]) and show that we can precisely solve these benchmarks. More than proposing an end-all community definition, we explain how studying the detailed properties of community definitions is important in order to validate that definitions do not have negative analytic properties. We urge that community definitions be separated from community detection algorithms and propose that community definitions be further evaluated by criteria such as these.
[ { "version": "v1", "created": "Mon, 14 Jan 2013 20:40:13 GMT" } ]
2013-01-15T00:00:00
[ [ "Ronhovde", "Richard K. Darst David R. Reichman Peter", "" ], [ "Nussinov", "Zohar", "" ] ]
TITLE: An edge density definition of overlapping and weighted graph communities ABSTRACT: Community detection in networks refers to the process of seeking strongly internally connected groups of nodes which are weakly externally connected. In this work, we introduce and study a community definition based on internal edge density. Beginning with the simple concept that edge density equals number of edges divided by maximal number of edges, we apply this definition to a variety of node and community arrangements to show that our definition yields sensible results. Our community definition is equivalent to that of the Absolute Potts Model community detection method (Phys. Rev. E 81, 046114 (2010)), and the performance of that method validates the usefulness of our definition across a wide variety of network types. We discuss how this definition can be extended to weighted, and multigraphs, and how the definition is capable of handling overlapping communities and local algorithms. We further validate our definition against the recently proposed Affiliation Graph Model (arXiv:1205.6228 [cs.SI]) and show that we can precisely solve these benchmarks. More than proposing an end-all community definition, we explain how studying the detailed properties of community definitions is important in order to validate that definitions do not have negative analytic properties. We urge that community definitions be separated from community detection algorithms and propose that community definitions be further evaluated by criteria such as these.
no_new_dataset
0.708389
1110.1980
Carmine Ventre
Paul W. Goldberg, Carmine Ventre
Using Lotteries to Approximate the Optimal Revenue
null
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
There has been much recent work on the revenue-raising properties of truthful mechanisms for selling goods to selfish bidders. Typically the revenue of a mechanism is compared against a benchmark (such as, the maximum revenue obtainable by an omniscient seller selling at a fixed price to at least two customers), with a view to understanding how much lower the mechanism's revenue is than the benchmark, in the worst case. We study this issue in the context of {\em lotteries}, where the seller may sell a probability of winning an item. We are interested in two general issues. Firstly, we aim at using the true optimum revenue as benchmark for our auctions. Secondly, we study the extent to which the expressive power resulting from lotteries, helps to improve the worst-case ratio. We study this in the well-known context of {\em digital goods}, where the production cost is zero. We show that in this scenario, collusion-resistant lotteries (these are lotteries for which no coalition of bidders exchanging side payments has an advantage in lying) are as powerful as truthful ones.
[ { "version": "v1", "created": "Mon, 10 Oct 2011 09:58:16 GMT" }, { "version": "v2", "created": "Fri, 11 Jan 2013 15:14:06 GMT" } ]
2013-01-14T00:00:00
[ [ "Goldberg", "Paul W.", "" ], [ "Ventre", "Carmine", "" ] ]
TITLE: Using Lotteries to Approximate the Optimal Revenue ABSTRACT: There has been much recent work on the revenue-raising properties of truthful mechanisms for selling goods to selfish bidders. Typically the revenue of a mechanism is compared against a benchmark (such as, the maximum revenue obtainable by an omniscient seller selling at a fixed price to at least two customers), with a view to understanding how much lower the mechanism's revenue is than the benchmark, in the worst case. We study this issue in the context of {\em lotteries}, where the seller may sell a probability of winning an item. We are interested in two general issues. Firstly, we aim at using the true optimum revenue as benchmark for our auctions. Secondly, we study the extent to which the expressive power resulting from lotteries, helps to improve the worst-case ratio. We study this in the well-known context of {\em digital goods}, where the production cost is zero. We show that in this scenario, collusion-resistant lotteries (these are lotteries for which no coalition of bidders exchanging side payments has an advantage in lying) are as powerful as truthful ones.
no_new_dataset
0.706418
1301.2283
Tomas Kocka
Tomas Kocka, Robert Castelo
Improved learning of Bayesian networks
Appears in Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence (UAI2001)
null
null
UAI-P-2001-PG-269-276
cs.LG cs.AI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The search space of Bayesian Network structures is usually defined as Acyclic Directed Graphs (DAGs) and the search is done by local transformations of DAGs. But the space of Bayesian Networks is ordered by DAG Markov model inclusion and it is natural to consider that a good search policy should take this into account. First attempt to do this (Chickering 1996) was using equivalence classes of DAGs instead of DAGs itself. This approach produces better results but it is significantly slower. We present a compromise between these two approaches. It uses DAGs to search the space in such a way that the ordering by inclusion is taken into account. This is achieved by repetitive usage of local moves within the equivalence class of DAGs. We show that this new approach produces better results than the original DAGs approach without substantial change in time complexity. We present empirical results, within the framework of heuristic search and Markov Chain Monte Carlo, provided through the Alarm dataset.
[ { "version": "v1", "created": "Thu, 10 Jan 2013 16:24:32 GMT" } ]
2013-01-14T00:00:00
[ [ "Kocka", "Tomas", "" ], [ "Castelo", "Robert", "" ] ]
TITLE: Improved learning of Bayesian networks ABSTRACT: The search space of Bayesian Network structures is usually defined as Acyclic Directed Graphs (DAGs) and the search is done by local transformations of DAGs. But the space of Bayesian Networks is ordered by DAG Markov model inclusion and it is natural to consider that a good search policy should take this into account. First attempt to do this (Chickering 1996) was using equivalence classes of DAGs instead of DAGs itself. This approach produces better results but it is significantly slower. We present a compromise between these two approaches. It uses DAGs to search the space in such a way that the ordering by inclusion is taken into account. This is achieved by repetitive usage of local moves within the equivalence class of DAGs. We show that this new approach produces better results than the original DAGs approach without substantial change in time complexity. We present empirical results, within the framework of heuristic search and Markov Chain Monte Carlo, provided through the Alarm dataset.
no_new_dataset
0.7114
1301.2375
Jianxin Li
Jianxin Li, Chengfei Liu, Liang Yao and Jeffrey Xu Yu
Context-based Diversification for Keyword Queries over XML Data
null
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
While keyword query empowers ordinary users to search vast amount of data, the ambiguity of keyword query makes it difficult to effectively answer keyword queries, especially for short and vague keyword queries. To address this challenging problem, in this paper we propose an approach that automatically diversifies XML keyword search based on its different contexts in the XML data. Given a short and vague keyword query and XML data to be searched, we firstly derive keyword search candidates of the query by a classifical feature selection model. And then, we design an effective XML keyword search diversification model to measure the quality of each candidate. After that, three efficient algorithms are proposed to evaluate the possible generated query candidates representing the diversified search intentions, from which we can find and return top-$k$ qualified query candidates that are most relevant to the given keyword query while they can cover maximal number of distinct results.At last, a comprehensive evaluation on real and synthetic datasets demonstrates the effectiveness of our proposed diversification model and the efficiency of our algorithms.
[ { "version": "v1", "created": "Fri, 11 Jan 2013 01:33:50 GMT" } ]
2013-01-14T00:00:00
[ [ "Li", "Jianxin", "" ], [ "Liu", "Chengfei", "" ], [ "Yao", "Liang", "" ], [ "Yu", "Jeffrey Xu", "" ] ]
TITLE: Context-based Diversification for Keyword Queries over XML Data ABSTRACT: While keyword query empowers ordinary users to search vast amount of data, the ambiguity of keyword query makes it difficult to effectively answer keyword queries, especially for short and vague keyword queries. To address this challenging problem, in this paper we propose an approach that automatically diversifies XML keyword search based on its different contexts in the XML data. Given a short and vague keyword query and XML data to be searched, we firstly derive keyword search candidates of the query by a classifical feature selection model. And then, we design an effective XML keyword search diversification model to measure the quality of each candidate. After that, three efficient algorithms are proposed to evaluate the possible generated query candidates representing the diversified search intentions, from which we can find and return top-$k$ qualified query candidates that are most relevant to the given keyword query while they can cover maximal number of distinct results.At last, a comprehensive evaluation on real and synthetic datasets demonstrates the effectiveness of our proposed diversification model and the efficiency of our algorithms.
no_new_dataset
0.712845
1301.2378
Jianxin Li
Jianxin Li, Chengfei Liu, Liang Yao, Jeffrey Xu Yu and Rui Zhou
Query-driven Frequent Co-occurring Term Extraction over Relational Data using MapReduce
null
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we study how to efficiently compute \textit{frequent co-occurring terms} (FCT) in the results of a keyword query in parallel using the popular MapReduce framework. Taking as input a keyword query q and an integer k, an FCT query reports the k terms that are not in q, but appear most frequently in the results of the keyword query q over multiple joined relations. The returned terms of FCT search can be used to do query expansion and query refinement for traditional keyword search. Different from the method of FCT search in a single platform, our proposed approach can efficiently answer a FCT query using the MapReduce Paradigm without pre-computing the results of the original keyword query, which is run in parallel platform. In this work, we can output the final FCT search results by two MapReduce jobs: the first is to extract the statistical information of the data; and the second is to calculate the total frequency of each term based on the output of the first job. At the two MapReduce jobs, we would guarantee the load balance of mappers and the computational balance of reducers as much as possible. Analytical and experimental evaluations demonstrate the efficiency and scalability of our proposed approach using TPC-H benchmark datasets with different sizes.
[ { "version": "v1", "created": "Fri, 11 Jan 2013 01:55:10 GMT" } ]
2013-01-14T00:00:00
[ [ "Li", "Jianxin", "" ], [ "Liu", "Chengfei", "" ], [ "Yao", "Liang", "" ], [ "Yu", "Jeffrey Xu", "" ], [ "Zhou", "Rui", "" ] ]
TITLE: Query-driven Frequent Co-occurring Term Extraction over Relational Data using MapReduce ABSTRACT: In this paper we study how to efficiently compute \textit{frequent co-occurring terms} (FCT) in the results of a keyword query in parallel using the popular MapReduce framework. Taking as input a keyword query q and an integer k, an FCT query reports the k terms that are not in q, but appear most frequently in the results of the keyword query q over multiple joined relations. The returned terms of FCT search can be used to do query expansion and query refinement for traditional keyword search. Different from the method of FCT search in a single platform, our proposed approach can efficiently answer a FCT query using the MapReduce Paradigm without pre-computing the results of the original keyword query, which is run in parallel platform. In this work, we can output the final FCT search results by two MapReduce jobs: the first is to extract the statistical information of the data; and the second is to calculate the total frequency of each term based on the output of the first job. At the two MapReduce jobs, we would guarantee the load balance of mappers and the computational balance of reducers as much as possible. Analytical and experimental evaluations demonstrate the efficiency and scalability of our proposed approach using TPC-H benchmark datasets with different sizes.
no_new_dataset
0.709252
1301.2012
Srivatsan Laxman
Srivatsan Laxman, Sushil Mittal and Ramarathnam Venkatesan
Error Correction in Learning using SVMs
null
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper is concerned with learning binary classifiers under adversarial label-noise. We introduce the problem of error-correction in learning where the goal is to recover the original clean data from a label-manipulated version of it, given (i) no constraints on the adversary other than an upper-bound on the number of errors, and (ii) some regularity properties for the original data. We present a simple and practical error-correction algorithm called SubSVMs that learns individual SVMs on several small-size (log-size), class-balanced, random subsets of the data and then reclassifies the training points using a majority vote. Our analysis reveals the need for the two main ingredients of SubSVMs, namely class-balanced sampling and subsampled bagging. Experimental results on synthetic as well as benchmark UCI data demonstrate the effectiveness of our approach. In addition to noise-tolerance, log-size subsampled bagging also yields significant run-time benefits over standard SVMs.
[ { "version": "v1", "created": "Thu, 10 Jan 2013 00:47:21 GMT" } ]
2013-01-11T00:00:00
[ [ "Laxman", "Srivatsan", "" ], [ "Mittal", "Sushil", "" ], [ "Venkatesan", "Ramarathnam", "" ] ]
TITLE: Error Correction in Learning using SVMs ABSTRACT: This paper is concerned with learning binary classifiers under adversarial label-noise. We introduce the problem of error-correction in learning where the goal is to recover the original clean data from a label-manipulated version of it, given (i) no constraints on the adversary other than an upper-bound on the number of errors, and (ii) some regularity properties for the original data. We present a simple and practical error-correction algorithm called SubSVMs that learns individual SVMs on several small-size (log-size), class-balanced, random subsets of the data and then reclassifies the training points using a majority vote. Our analysis reveals the need for the two main ingredients of SubSVMs, namely class-balanced sampling and subsampled bagging. Experimental results on synthetic as well as benchmark UCI data demonstrate the effectiveness of our approach. In addition to noise-tolerance, log-size subsampled bagging also yields significant run-time benefits over standard SVMs.
no_new_dataset
0.709044
1301.2115
Krikamol Muandet
Krikamol Muandet, David Balduzzi, Bernhard Sch\"olkopf
Domain Generalization via Invariant Feature Representation
The 30th International Conference on Machine Learning (ICML 2013)
null
null
null
stat.ML cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice.
[ { "version": "v1", "created": "Thu, 10 Jan 2013 13:29:17 GMT" } ]
2013-01-11T00:00:00
[ [ "Muandet", "Krikamol", "" ], [ "Balduzzi", "David", "" ], [ "Schölkopf", "Bernhard", "" ] ]
TITLE: Domain Generalization via Invariant Feature Representation ABSTRACT: This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice.
no_new_dataset
0.709837
1301.1701
Meysam Mirzaee
Meysam Mirzaee and Soroush Akhlaghi
Secrecy Capacity of Two-Hop Relay Assisted Wiretap Channels
null
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Incorporating the physical layer characteristics to secure communications has received considerable attention in recent years. Moreover, cooperation with some nodes of network can give benefits of multiple-antenna systems, increasing the secrecy capacity of such channels. In this paper, we consider cooperative wiretap channel with the help of an Amplify and Forward (AF) relay to transmit confidential messages from source to legitimate receiver in the presence of an eavesdropper. In this regard, the secrecy capacity of AF relying is derived, assuming the relay is subject to a peak power constraint. To this end, an achievable secrecy rate for Gaussian input is evaluated through solving a non-convex optimization problem. Then, it is proved that any rates greater than this secrecy rate is not achievable. To do this, the capacity of a genie-aided channel as an upper bound for the secrecy capacity of the underlying channel is derived, showing this upper bound is equal to the computed achievable secrecy rate with Gaussian input. Accordingly, the corresponding secrecy capacity is compared to the Decode and Forward (DF) strategy which is served as the benchmark in the current work.
[ { "version": "v1", "created": "Tue, 8 Jan 2013 21:46:21 GMT" } ]
2013-01-10T00:00:00
[ [ "Mirzaee", "Meysam", "" ], [ "Akhlaghi", "Soroush", "" ] ]
TITLE: Secrecy Capacity of Two-Hop Relay Assisted Wiretap Channels ABSTRACT: Incorporating the physical layer characteristics to secure communications has received considerable attention in recent years. Moreover, cooperation with some nodes of network can give benefits of multiple-antenna systems, increasing the secrecy capacity of such channels. In this paper, we consider cooperative wiretap channel with the help of an Amplify and Forward (AF) relay to transmit confidential messages from source to legitimate receiver in the presence of an eavesdropper. In this regard, the secrecy capacity of AF relying is derived, assuming the relay is subject to a peak power constraint. To this end, an achievable secrecy rate for Gaussian input is evaluated through solving a non-convex optimization problem. Then, it is proved that any rates greater than this secrecy rate is not achievable. To do this, the capacity of a genie-aided channel as an upper bound for the secrecy capacity of the underlying channel is derived, showing this upper bound is equal to the computed achievable secrecy rate with Gaussian input. Accordingly, the corresponding secrecy capacity is compared to the Decode and Forward (DF) strategy which is served as the benchmark in the current work.
no_new_dataset
0.710584
1301.1722
Andrea Montanari
Yash Deshpande and Andrea Montanari
Linear Bandits in High Dimension and Recommendation Systems
21 pages, 4 figures
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A large number of online services provide automated recommendations to help users to navigate through a large collection of items. New items (products, videos, songs, advertisements) are suggested on the basis of the user's past history and --when available-- her demographic profile. Recommendations have to satisfy the dual goal of helping the user to explore the space of available items, while allowing the system to probe the user's preferences. We model this trade-off using linearly parametrized multi-armed bandits, propose a policy and prove upper and lower bounds on the cumulative "reward" that coincide up to constants in the data poor (high-dimensional) regime. Prior work on linear bandits has focused on the data rich (low-dimensional) regime and used cumulative "risk" as the figure of merit. For this data rich regime, we provide a simple modification for our policy that achieves near-optimal risk performance under more restrictive assumptions on the geometry of the problem. We test (a variation of) the scheme used for establishing achievability on the Netflix and MovieLens datasets and obtain good agreement with the qualitative predictions of the theory we develop.
[ { "version": "v1", "created": "Tue, 8 Jan 2013 23:45:06 GMT" } ]
2013-01-10T00:00:00
[ [ "Deshpande", "Yash", "" ], [ "Montanari", "Andrea", "" ] ]
TITLE: Linear Bandits in High Dimension and Recommendation Systems ABSTRACT: A large number of online services provide automated recommendations to help users to navigate through a large collection of items. New items (products, videos, songs, advertisements) are suggested on the basis of the user's past history and --when available-- her demographic profile. Recommendations have to satisfy the dual goal of helping the user to explore the space of available items, while allowing the system to probe the user's preferences. We model this trade-off using linearly parametrized multi-armed bandits, propose a policy and prove upper and lower bounds on the cumulative "reward" that coincide up to constants in the data poor (high-dimensional) regime. Prior work on linear bandits has focused on the data rich (low-dimensional) regime and used cumulative "risk" as the figure of merit. For this data rich regime, we provide a simple modification for our policy that achieves near-optimal risk performance under more restrictive assumptions on the geometry of the problem. We test (a variation of) the scheme used for establishing achievability on the Netflix and MovieLens datasets and obtain good agreement with the qualitative predictions of the theory we develop.
no_new_dataset
0.70844
1208.3422
Zhixiang Eddie Xu
Zhixiang Xu, Kilian Q. Weinberger, Olivier Chapelle
Distance Metric Learning for Kernel Machines
null
null
null
null
stat.ML cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recent work in metric learning has significantly improved the state-of-the-art in k-nearest neighbor classification. Support vector machines (SVM), particularly with RBF kernels, are amongst the most popular classification algorithms that uses distance metrics to compare examples. This paper provides an empirical analysis of the efficacy of three of the most popular Mahalanobis metric learning algorithms as pre-processing for SVM training. We show that none of these algorithms generate metrics that lead to particularly satisfying improvements for SVM-RBF classification. As a remedy we introduce support vector metric learning (SVML), a novel algorithm that seamlessly combines the learning of a Mahalanobis metric with the training of the RBF-SVM parameters. We demonstrate the capabilities of SVML on nine benchmark data sets of varying sizes and difficulties. In our study, SVML outperforms all alternative state-of-the-art metric learning algorithms in terms of accuracy and establishes itself as a serious alternative to the standard Euclidean metric with model selection by cross validation.
[ { "version": "v1", "created": "Thu, 16 Aug 2012 17:16:18 GMT" }, { "version": "v2", "created": "Tue, 8 Jan 2013 20:26:55 GMT" } ]
2013-01-09T00:00:00
[ [ "Xu", "Zhixiang", "" ], [ "Weinberger", "Kilian Q.", "" ], [ "Chapelle", "Olivier", "" ] ]
TITLE: Distance Metric Learning for Kernel Machines ABSTRACT: Recent work in metric learning has significantly improved the state-of-the-art in k-nearest neighbor classification. Support vector machines (SVM), particularly with RBF kernels, are amongst the most popular classification algorithms that uses distance metrics to compare examples. This paper provides an empirical analysis of the efficacy of three of the most popular Mahalanobis metric learning algorithms as pre-processing for SVM training. We show that none of these algorithms generate metrics that lead to particularly satisfying improvements for SVM-RBF classification. As a remedy we introduce support vector metric learning (SVML), a novel algorithm that seamlessly combines the learning of a Mahalanobis metric with the training of the RBF-SVM parameters. We demonstrate the capabilities of SVML on nine benchmark data sets of varying sizes and difficulties. In our study, SVML outperforms all alternative state-of-the-art metric learning algorithms in terms of accuracy and establishes itself as a serious alternative to the standard Euclidean metric with model selection by cross validation.
no_new_dataset
0.71
1301.1502
Hannah Inbarani
N. Kalaiselvi, H. Hannah Inbarani
Fuzzy Soft Set Based Classification for Gene Expression Data
7 pages, IJSER Vol.3 Issue: 10 Oct 2012
null
null
null
cs.AI cs.CE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Classification is one of the major issues in Data Mining Research fields. The classification problems in medical area often classify medical dataset based on the result of medical diagnosis or description of medical treatment by the medical practitioner. This research work discusses the classification process of Gene Expression data for three different cancers which are breast cancer, lung cancer and leukemia cancer with two classes which are cancerous stage and non cancerous stage. We have applied a fuzzy soft set similarity based classifier to enhance the accuracy to predict the stages among cancer genes and the informative genes are selected by using Entopy filtering.
[ { "version": "v1", "created": "Tue, 8 Jan 2013 11:48:49 GMT" } ]
2013-01-09T00:00:00
[ [ "Kalaiselvi", "N.", "" ], [ "Inbarani", "H. Hannah", "" ] ]
TITLE: Fuzzy Soft Set Based Classification for Gene Expression Data ABSTRACT: Classification is one of the major issues in Data Mining Research fields. The classification problems in medical area often classify medical dataset based on the result of medical diagnosis or description of medical treatment by the medical practitioner. This research work discusses the classification process of Gene Expression data for three different cancers which are breast cancer, lung cancer and leukemia cancer with two classes which are cancerous stage and non cancerous stage. We have applied a fuzzy soft set similarity based classifier to enhance the accuracy to predict the stages among cancer genes and the informative genes are selected by using Entopy filtering.
no_new_dataset
0.714036
1104.4874
Jan Treibig
Jan Treibig, Georg Hager, and Gerhard Wellein
LIKWID: Lightweight Performance Tools
12 pages
null
10.1109/ICPPW.2010.38
null
cs.DC cs.PF
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Exploiting the performance of today's microprocessors requires intimate knowledge of the microarchitecture as well as an awareness of the ever-growing complexity in thread and cache topology. LIKWID is a set of command line utilities that addresses four key problems: Probing the thread and cache topology of a shared-memory node, enforcing thread-core affinity on a program, measuring performance counter metrics, and microbenchmarking for reliable upper performance bounds. Moreover, it includes a mpirun wrapper allowing for portable thread-core affinity in MPI and hybrid MPI/threaded applications. To demonstrate the capabilities of the tool set we show the influence of thread affinity on performance using the well-known OpenMP STREAM triad benchmark, use hardware counter tools to study the performance of a stencil code, and finally show how to detect bandwidth problems on ccNUMA-based compute nodes.
[ { "version": "v1", "created": "Tue, 26 Apr 2011 09:43:46 GMT" }, { "version": "v2", "created": "Mon, 7 Jan 2013 09:03:29 GMT" } ]
2013-01-08T00:00:00
[ [ "Treibig", "Jan", "" ], [ "Hager", "Georg", "" ], [ "Wellein", "Gerhard", "" ] ]
TITLE: LIKWID: Lightweight Performance Tools ABSTRACT: Exploiting the performance of today's microprocessors requires intimate knowledge of the microarchitecture as well as an awareness of the ever-growing complexity in thread and cache topology. LIKWID is a set of command line utilities that addresses four key problems: Probing the thread and cache topology of a shared-memory node, enforcing thread-core affinity on a program, measuring performance counter metrics, and microbenchmarking for reliable upper performance bounds. Moreover, it includes a mpirun wrapper allowing for portable thread-core affinity in MPI and hybrid MPI/threaded applications. To demonstrate the capabilities of the tool set we show the influence of thread affinity on performance using the well-known OpenMP STREAM triad benchmark, use hardware counter tools to study the performance of a stencil code, and finally show how to detect bandwidth problems on ccNUMA-based compute nodes.
no_new_dataset
0.703753
1208.0959
Misha Denil
Misha Denil and Nando de Freitas
Recklessly Approximate Sparse Coding
null
null
null
null
cs.LG cs.CV stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
It has recently been observed that certain extremely simple feature encoding techniques are able to achieve state of the art performance on several standard image classification benchmarks including deep belief networks, convolutional nets, factored RBMs, mcRBMs, convolutional RBMs, sparse autoencoders and several others. Moreover, these "triangle" or "soft threshold" encodings are ex- tremely efficient to compute. Several intuitive arguments have been put forward to explain this remarkable performance, yet no mathematical justification has been offered. The main result of this report is to show that these features are realized as an approximate solution to the a non-negative sparse coding problem. Using this connection we describe several variants of the soft threshold features and demonstrate their effectiveness on two image classification benchmark tasks.
[ { "version": "v1", "created": "Sat, 4 Aug 2012 21:48:52 GMT" }, { "version": "v2", "created": "Sun, 6 Jan 2013 19:00:48 GMT" } ]
2013-01-08T00:00:00
[ [ "Denil", "Misha", "" ], [ "de Freitas", "Nando", "" ] ]
TITLE: Recklessly Approximate Sparse Coding ABSTRACT: It has recently been observed that certain extremely simple feature encoding techniques are able to achieve state of the art performance on several standard image classification benchmarks including deep belief networks, convolutional nets, factored RBMs, mcRBMs, convolutional RBMs, sparse autoencoders and several others. Moreover, these "triangle" or "soft threshold" encodings are ex- tremely efficient to compute. Several intuitive arguments have been put forward to explain this remarkable performance, yet no mathematical justification has been offered. The main result of this report is to show that these features are realized as an approximate solution to the a non-negative sparse coding problem. Using this connection we describe several variants of the soft threshold features and demonstrate their effectiveness on two image classification benchmark tasks.
no_new_dataset
0.709007
1212.4775
Mario Frank
Mario Frank, Joachim M. Buhmann, David Basin
Role Mining with Probabilistic Models
accepted for publication at ACM Transactions on Information and System Security (TISSEC)
null
null
null
cs.CR cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Role mining tackles the problem of finding a role-based access control (RBAC) configuration, given an access-control matrix assigning users to access permissions as input. Most role mining approaches work by constructing a large set of candidate roles and use a greedy selection strategy to iteratively pick a small subset such that the differences between the resulting RBAC configuration and the access control matrix are minimized. In this paper, we advocate an alternative approach that recasts role mining as an inference problem rather than a lossy compression problem. Instead of using combinatorial algorithms to minimize the number of roles needed to represent the access-control matrix, we derive probabilistic models to learn the RBAC configuration that most likely underlies the given matrix. Our models are generative in that they reflect the way that permissions are assigned to users in a given RBAC configuration. We additionally model how user-permission assignments that conflict with an RBAC configuration emerge and we investigate the influence of constraints on role hierarchies and on the number of assignments. In experiments with access-control matrices from real-world enterprises, we compare our proposed models with other role mining methods. Our results show that our probabilistic models infer roles that generalize well to new system users for a wide variety of data, while other models' generalization abilities depend on the dataset given.
[ { "version": "v1", "created": "Wed, 19 Dec 2012 18:12:34 GMT" }, { "version": "v2", "created": "Thu, 3 Jan 2013 17:27:55 GMT" }, { "version": "v3", "created": "Fri, 4 Jan 2013 22:24:15 GMT" } ]
2013-01-08T00:00:00
[ [ "Frank", "Mario", "" ], [ "Buhmann", "Joachim M.", "" ], [ "Basin", "David", "" ] ]
TITLE: Role Mining with Probabilistic Models ABSTRACT: Role mining tackles the problem of finding a role-based access control (RBAC) configuration, given an access-control matrix assigning users to access permissions as input. Most role mining approaches work by constructing a large set of candidate roles and use a greedy selection strategy to iteratively pick a small subset such that the differences between the resulting RBAC configuration and the access control matrix are minimized. In this paper, we advocate an alternative approach that recasts role mining as an inference problem rather than a lossy compression problem. Instead of using combinatorial algorithms to minimize the number of roles needed to represent the access-control matrix, we derive probabilistic models to learn the RBAC configuration that most likely underlies the given matrix. Our models are generative in that they reflect the way that permissions are assigned to users in a given RBAC configuration. We additionally model how user-permission assignments that conflict with an RBAC configuration emerge and we investigate the influence of constraints on role hierarchies and on the number of assignments. In experiments with access-control matrices from real-world enterprises, we compare our proposed models with other role mining methods. Our results show that our probabilistic models infer roles that generalize well to new system users for a wide variety of data, while other models' generalization abilities depend on the dataset given.
no_new_dataset
0.709982
1301.0930
Saptarshi Das
Sayan Saha, Saptarshi Das, Anindya Pakhira, Sumit Mukherjee, Indranil Pan
Comparative Studies on Decentralized Multiloop PID Controller Design Using Evolutionary Algorithms
6 pages, 9 figures
Engineering and Systems (SCES), 2012 Students Conference on; Allahabad, March 2012
10.1109/SCES.2012.6199122
null
cs.SY cs.NE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Decentralized PID controllers have been designed in this paper for simultaneous tracking of individual process variables in multivariable systems under step reference input. The controller design framework takes into account the minimization of a weighted sum of Integral of Time multiplied Squared Error (ITSE) and Integral of Squared Controller Output (ISCO) so as to balance the overall tracking errors for the process variables and required variation in the corresponding manipulated variables. Decentralized PID gains are tuned using three popular Evolutionary Algorithms (EAs) viz. Genetic Algorithm (GA), Evolutionary Strategy (ES) and Cultural Algorithm (CA). Credible simulation comparisons have been reported for four benchmark 2x2 multivariable processes.
[ { "version": "v1", "created": "Sat, 5 Jan 2013 18:45:01 GMT" } ]
2013-01-08T00:00:00
[ [ "Saha", "Sayan", "" ], [ "Das", "Saptarshi", "" ], [ "Pakhira", "Anindya", "" ], [ "Mukherjee", "Sumit", "" ], [ "Pan", "Indranil", "" ] ]
TITLE: Comparative Studies on Decentralized Multiloop PID Controller Design Using Evolutionary Algorithms ABSTRACT: Decentralized PID controllers have been designed in this paper for simultaneous tracking of individual process variables in multivariable systems under step reference input. The controller design framework takes into account the minimization of a weighted sum of Integral of Time multiplied Squared Error (ITSE) and Integral of Squared Controller Output (ISCO) so as to balance the overall tracking errors for the process variables and required variation in the corresponding manipulated variables. Decentralized PID gains are tuned using three popular Evolutionary Algorithms (EAs) viz. Genetic Algorithm (GA), Evolutionary Strategy (ES) and Cultural Algorithm (CA). Credible simulation comparisons have been reported for four benchmark 2x2 multivariable processes.
no_new_dataset
0.709868
1301.0561
David Maxwell Chickering
David Maxwell Chickering, Christopher Meek
Finding Optimal Bayesian Networks
Appears in Proceedings of the Eighteenth Conference on Uncertainty in Artificial Intelligence (UAI2002)
null
null
UAI-P-2002-PG-94-102
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we derive optimality results for greedy Bayesian-network search algorithms that perform single-edge modifications at each step and use asymptotically consistent scoring criteria. Our results extend those of Meek (1997) and Chickering (2002), who demonstrate that in the limit of large datasets, if the generative distribution is perfect with respect to a DAG defined over the observable variables, such search algorithms will identify this optimal (i.e. generative) DAG model. We relax their assumption about the generative distribution, and assume only that this distribution satisfies the {em composition property} over the observable variables, which is a more realistic assumption for real domains. Under this assumption, we guarantee that the search algorithms identify an {em inclusion-optimal} model; that is, a model that (1) contains the generative distribution and (2) has no sub-model that contains this distribution. In addition, we show that the composition property is guaranteed to hold whenever the dependence relationships in the generative distribution can be characterized by paths between singleton elements in some generative graphical model (e.g. a DAG, a chain graph, or a Markov network) even when the generative model includes unobserved variables, and even when the observed data is subject to selection bias.
[ { "version": "v1", "created": "Wed, 12 Dec 2012 15:55:46 GMT" } ]
2013-01-07T00:00:00
[ [ "Chickering", "David Maxwell", "" ], [ "Meek", "Christopher", "" ] ]
TITLE: Finding Optimal Bayesian Networks ABSTRACT: In this paper, we derive optimality results for greedy Bayesian-network search algorithms that perform single-edge modifications at each step and use asymptotically consistent scoring criteria. Our results extend those of Meek (1997) and Chickering (2002), who demonstrate that in the limit of large datasets, if the generative distribution is perfect with respect to a DAG defined over the observable variables, such search algorithms will identify this optimal (i.e. generative) DAG model. We relax their assumption about the generative distribution, and assume only that this distribution satisfies the {em composition property} over the observable variables, which is a more realistic assumption for real domains. Under this assumption, we guarantee that the search algorithms identify an {em inclusion-optimal} model; that is, a model that (1) contains the generative distribution and (2) has no sub-model that contains this distribution. In addition, we show that the composition property is guaranteed to hold whenever the dependence relationships in the generative distribution can be characterized by paths between singleton elements in some generative graphical model (e.g. a DAG, a chain graph, or a Markov network) even when the generative model includes unobserved variables, and even when the observed data is subject to selection bias.
no_new_dataset
0.7114
1301.0702
Sundeep Prabhakar Chepuri
Sundeep Prabhakar Chepuri, Geert Leus, and Alle-Jan van der Veen
Joint localization and clock synchronization for wireless sensor networks
5 pages, 4 figures, to appear in proc. of Asilomar 2012
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A fully-asynchronous network with one target sensor and a few anchors (nodes with known locations) is considered. Localization and synchronization are traditionally treated as two separate problems. In this paper, localization and synchronization is studied under a unified framework. We present a new model in which time-stamps obtained either via two-way communication between the nodes or with a broadcast based protocol can be used in a simple estimator based on least-squares (LS) to jointly estimate the position of the target node as well as all the unknown clock-skews and clock-offsets. The Cram\'er-Rao lower bound (CRLB) is derived for the considered problem and is used as a benchmark to analyze the performance of the proposed estimator.
[ { "version": "v1", "created": "Fri, 4 Jan 2013 10:41:58 GMT" } ]
2013-01-07T00:00:00
[ [ "Chepuri", "Sundeep Prabhakar", "" ], [ "Leus", "Geert", "" ], [ "van der Veen", "Alle-Jan", "" ] ]
TITLE: Joint localization and clock synchronization for wireless sensor networks ABSTRACT: A fully-asynchronous network with one target sensor and a few anchors (nodes with known locations) is considered. Localization and synchronization are traditionally treated as two separate problems. In this paper, localization and synchronization is studied under a unified framework. We present a new model in which time-stamps obtained either via two-way communication between the nodes or with a broadcast based protocol can be used in a simple estimator based on least-squares (LS) to jointly estimate the position of the target node as well as all the unknown clock-skews and clock-offsets. The Cram\'er-Rao lower bound (CRLB) is derived for the considered problem and is used as a benchmark to analyze the performance of the proposed estimator.
no_new_dataset
0.709837
1301.0432
Fahad Mahmood Mr
F. Mahmood, F. Kunwar
A Self-Organizing Neural Scheme for Door Detection in Different Environments
Page No 13-18, 7 figures, Published with International Journal of Computer Applications (IJCA)
International Journal of Computer Applications 60(9):13-18, 2012
10.5120/9719-3679
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Doors are important landmarks for indoor mobile robot navigation and also assist blind people to independently access unfamiliar buildings. Most existing algorithms of door detection are limited to work for familiar environments because of restricted assumptions about color, texture and shape. In this paper we propose a novel approach which employs feature based classification and uses the Kohonen Self-Organizing Map (SOM) for the purpose of door detection. Generic and stable features are used for the training of SOM that increase the performance significantly: concavity, bottom-edge intensity profile and door edges. To validate the robustness and generalizability of our method, we collected a large dataset of real world door images from a variety of environments and different lighting conditions. The algorithm achieves more than 95% detection which demonstrates that our door detection method is generic and robust with variations of color, texture, occlusions, lighting condition, scales, and viewpoints.
[ { "version": "v1", "created": "Thu, 3 Jan 2013 12:04:28 GMT" } ]
2013-01-04T00:00:00
[ [ "Mahmood", "F.", "" ], [ "Kunwar", "F.", "" ] ]
TITLE: A Self-Organizing Neural Scheme for Door Detection in Different Environments ABSTRACT: Doors are important landmarks for indoor mobile robot navigation and also assist blind people to independently access unfamiliar buildings. Most existing algorithms of door detection are limited to work for familiar environments because of restricted assumptions about color, texture and shape. In this paper we propose a novel approach which employs feature based classification and uses the Kohonen Self-Organizing Map (SOM) for the purpose of door detection. Generic and stable features are used for the training of SOM that increase the performance significantly: concavity, bottom-edge intensity profile and door edges. To validate the robustness and generalizability of our method, we collected a large dataset of real world door images from a variety of environments and different lighting conditions. The algorithm achieves more than 95% detection which demonstrates that our door detection method is generic and robust with variations of color, texture, occlusions, lighting condition, scales, and viewpoints.
new_dataset
0.710446
0911.2942
Chris Giannella
Chris Giannella, Kun Liu, Hillol Kargupta
Breaching Euclidean Distance-Preserving Data Perturbation Using Few Known Inputs
This is a major revision accounting for journal peer-review. Changes include: removal of known sample attack, more citations added, an empirical comparison against the algorithm of Kaplan et al. added
Data & Knowledge Engineering 83, pages 93-110, 2013
null
null
cs.DB cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We examine Euclidean distance-preserving data perturbation as a tool for privacy-preserving data mining. Such perturbations allow many important data mining algorithms e.g. hierarchical and k-means clustering), with only minor modification, to be applied to the perturbed data and produce exactly the same results as if applied to the original data. However, the issue of how well the privacy of the original data is preserved needs careful study. We engage in this study by assuming the role of an attacker armed with a small set of known original data tuples (inputs). Little work has been done examining this kind of attack when the number of known original tuples is less than the number of data dimensions. We focus on this important case, develop and rigorously analyze an attack that utilizes any number of known original tuples. The approach allows the attacker to estimate the original data tuple associated with each perturbed tuple and calculate the probability that the estimation results in a privacy breach. On a real 16-dimensional dataset, we show that the attacker, with 4 known original tuples, can estimate an original unknown tuple with less than 7% error with probability exceeding 0.8.
[ { "version": "v1", "created": "Mon, 16 Nov 2009 02:51:37 GMT" }, { "version": "v2", "created": "Wed, 2 Jan 2013 15:49:10 GMT" } ]
2013-01-03T00:00:00
[ [ "Giannella", "Chris", "" ], [ "Liu", "Kun", "" ], [ "Kargupta", "Hillol", "" ] ]
TITLE: Breaching Euclidean Distance-Preserving Data Perturbation Using Few Known Inputs ABSTRACT: We examine Euclidean distance-preserving data perturbation as a tool for privacy-preserving data mining. Such perturbations allow many important data mining algorithms e.g. hierarchical and k-means clustering), with only minor modification, to be applied to the perturbed data and produce exactly the same results as if applied to the original data. However, the issue of how well the privacy of the original data is preserved needs careful study. We engage in this study by assuming the role of an attacker armed with a small set of known original data tuples (inputs). Little work has been done examining this kind of attack when the number of known original tuples is less than the number of data dimensions. We focus on this important case, develop and rigorously analyze an attack that utilizes any number of known original tuples. The approach allows the attacker to estimate the original data tuple associated with each perturbed tuple and calculate the probability that the estimation results in a privacy breach. On a real 16-dimensional dataset, we show that the attacker, with 4 known original tuples, can estimate an original unknown tuple with less than 7% error with probability exceeding 0.8.
no_new_dataset
0.710201
1212.5841
Andrei Zinovyev Dr.
Andrei Zinovyev and Evgeny Mirkes
Data complexity measured by principal graphs
Computers and Mathematics with Applications, in press
null
10.1016/j.camwa.2012.12.009
null
cs.LG cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
How to measure the complexity of a finite set of vectors embedded in a multidimensional space? This is a non-trivial question which can be approached in many different ways. Here we suggest a set of data complexity measures using universal approximators, principal cubic complexes. Principal cubic complexes generalise the notion of principal manifolds for datasets with non-trivial topologies. The type of the principal cubic complex is determined by its dimension and a grammar of elementary graph transformations. The simplest grammar produces principal trees. We introduce three natural types of data complexity: 1) geometric (deviation of the data's approximator from some "idealized" configuration, such as deviation from harmonicity); 2) structural (how many elements of a principal graph are needed to approximate the data), and 3) construction complexity (how many applications of elementary graph transformations are needed to construct the principal object starting from the simplest one). We compute these measures for several simulated and real-life data distributions and show them in the "accuracy-complexity" plots, helping to optimize the accuracy/complexity ratio. We discuss various issues connected with measuring data complexity. Software for computing data complexity measures from principal cubic complexes is provided as well.
[ { "version": "v1", "created": "Sun, 23 Dec 2012 23:20:14 GMT" }, { "version": "v2", "created": "Wed, 2 Jan 2013 00:00:40 GMT" } ]
2013-01-03T00:00:00
[ [ "Zinovyev", "Andrei", "" ], [ "Mirkes", "Evgeny", "" ] ]
TITLE: Data complexity measured by principal graphs ABSTRACT: How to measure the complexity of a finite set of vectors embedded in a multidimensional space? This is a non-trivial question which can be approached in many different ways. Here we suggest a set of data complexity measures using universal approximators, principal cubic complexes. Principal cubic complexes generalise the notion of principal manifolds for datasets with non-trivial topologies. The type of the principal cubic complex is determined by its dimension and a grammar of elementary graph transformations. The simplest grammar produces principal trees. We introduce three natural types of data complexity: 1) geometric (deviation of the data's approximator from some "idealized" configuration, such as deviation from harmonicity); 2) structural (how many elements of a principal graph are needed to approximate the data), and 3) construction complexity (how many applications of elementary graph transformations are needed to construct the principal object starting from the simplest one). We compute these measures for several simulated and real-life data distributions and show them in the "accuracy-complexity" plots, helping to optimize the accuracy/complexity ratio. We discuss various issues connected with measuring data complexity. Software for computing data complexity measures from principal cubic complexes is provided as well.
no_new_dataset
0.710057
1212.6316
Nathalie Villa-Vialaneix
Madalina Olteanu (SAMM), Nathalie Villa-Vialaneix (SAMM), Marie Cottrell (SAMM)
On-line relational SOM for dissimilarity data
WSOM 2012, Santiago : Chile (2012)
null
null
null
stat.ML cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In some applications and in order to address real world situations better, data may be more complex than simple vectors. In some examples, they can be known through their pairwise dissimilarities only. Several variants of the Self Organizing Map algorithm were introduced to generalize the original algorithm to this framework. Whereas median SOM is based on a rough representation of the prototypes, relational SOM allows representing these prototypes by a virtual combination of all elements in the data set. However, this latter approach suffers from two main drawbacks. First, its complexity can be large. Second, only a batch version of this algorithm has been studied so far and it often provides results having a bad topographic organization. In this article, an on-line version of relational SOM is described and justified. The algorithm is tested on several datasets, including categorical data and graphs, and compared with the batch version and with other SOM algorithms for non vector data.
[ { "version": "v1", "created": "Thu, 27 Dec 2012 07:07:06 GMT" } ]
2013-01-03T00:00:00
[ [ "Olteanu", "Madalina", "", "SAMM" ], [ "Villa-Vialaneix", "Nathalie", "", "SAMM" ], [ "Cottrell", "Marie", "", "SAMM" ] ]
TITLE: On-line relational SOM for dissimilarity data ABSTRACT: In some applications and in order to address real world situations better, data may be more complex than simple vectors. In some examples, they can be known through their pairwise dissimilarities only. Several variants of the Self Organizing Map algorithm were introduced to generalize the original algorithm to this framework. Whereas median SOM is based on a rough representation of the prototypes, relational SOM allows representing these prototypes by a virtual combination of all elements in the data set. However, this latter approach suffers from two main drawbacks. First, its complexity can be large. Second, only a batch version of this algorithm has been studied so far and it often provides results having a bad topographic organization. In this article, an on-line version of relational SOM is described and justified. The algorithm is tested on several datasets, including categorical data and graphs, and compared with the batch version and with other SOM algorithms for non vector data.
no_new_dataset
0.710848
1301.0082
F. Ozgur Catak
F. Ozgur Catak and M. Erdal Balaban
CloudSVM : Training an SVM Classifier in Cloud Computing Systems
13 pages
null
null
null
cs.LG cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In conventional method, distributed support vector machines (SVM) algorithms are trained over pre-configured intranet/internet environments to find out an optimal classifier. These methods are very complicated and costly for large datasets. Hence, we propose a method that is referred as the Cloud SVM training mechanism (CloudSVM) in a cloud computing environment with MapReduce technique for distributed machine learning applications. Accordingly, (i) SVM algorithm is trained in distributed cloud storage servers that work concurrently; (ii) merge all support vectors in every trained cloud node; and (iii) iterate these two steps until the SVM converges to the optimal classifier function. Large scale data sets are not possible to train using SVM algorithm on a single computer. The results of this study are important for training of large scale data sets for machine learning applications. We provided that iterative training of splitted data set in cloud computing environment using SVM will converge to a global optimal classifier in finite iteration size.
[ { "version": "v1", "created": "Tue, 1 Jan 2013 13:20:27 GMT" } ]
2013-01-03T00:00:00
[ [ "Catak", "F. Ozgur", "" ], [ "Balaban", "M. Erdal", "" ] ]
TITLE: CloudSVM : Training an SVM Classifier in Cloud Computing Systems ABSTRACT: In conventional method, distributed support vector machines (SVM) algorithms are trained over pre-configured intranet/internet environments to find out an optimal classifier. These methods are very complicated and costly for large datasets. Hence, we propose a method that is referred as the Cloud SVM training mechanism (CloudSVM) in a cloud computing environment with MapReduce technique for distributed machine learning applications. Accordingly, (i) SVM algorithm is trained in distributed cloud storage servers that work concurrently; (ii) merge all support vectors in every trained cloud node; and (iii) iterate these two steps until the SVM converges to the optimal classifier function. Large scale data sets are not possible to train using SVM algorithm on a single computer. The results of this study are important for training of large scale data sets for machine learning applications. We provided that iterative training of splitted data set in cloud computing environment using SVM will converge to a global optimal classifier in finite iteration size.
no_new_dataset
0.712801
1301.0289
Aaditya Prakash
Aaditya Prakash
Reconstructing Self Organizing Maps as Spider Graphs for better visual interpretation of large unstructured datasets
9 pages, 8 figures
null
null
null
cs.GR stat.ML
http://creativecommons.org/licenses/by/3.0/
Self-Organizing Maps (SOM) are popular unsupervised artificial neural network used to reduce dimensions and visualize data. Visual interpretation from Self-Organizing Maps (SOM) has been limited due to grid approach of data representation, which makes inter-scenario analysis impossible. The paper proposes a new way to structure SOM. This model reconstructs SOM to show strength between variables as the threads of a cobweb and illuminate inter-scenario analysis. While Radar Graphs are very crude representation of spider web, this model uses more lively and realistic cobweb representation to take into account the difference in strength and length of threads. This model allows for visualization of highly unstructured dataset with large number of dimensions, common in Bigdata sources.
[ { "version": "v1", "created": "Mon, 24 Dec 2012 17:10:28 GMT" } ]
2013-01-03T00:00:00
[ [ "Prakash", "Aaditya", "" ] ]
TITLE: Reconstructing Self Organizing Maps as Spider Graphs for better visual interpretation of large unstructured datasets ABSTRACT: Self-Organizing Maps (SOM) are popular unsupervised artificial neural network used to reduce dimensions and visualize data. Visual interpretation from Self-Organizing Maps (SOM) has been limited due to grid approach of data representation, which makes inter-scenario analysis impossible. The paper proposes a new way to structure SOM. This model reconstructs SOM to show strength between variables as the threads of a cobweb and illuminate inter-scenario analysis. While Radar Graphs are very crude representation of spider web, this model uses more lively and realistic cobweb representation to take into account the difference in strength and length of threads. This model allows for visualization of highly unstructured dataset with large number of dimensions, common in Bigdata sources.
no_new_dataset
0.713032
1202.5909
Rodrigo Aldecoa
Rodrigo Aldecoa and Ignacio Mar\'in
Closed benchmarks for network community structure characterization
18 pages, 5 figures. Available at http://pre.aps.org/abstract/PRE/v85/i2/e026109
Phys. Rev. E 85, 026109 (2012)
10.1103/PhysRevE.85.026109
null
physics.soc-ph cond-mat.stat-mech cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Characterizing the community structure of complex networks is a key challenge in many scientific fields. Very diverse algorithms and methods have been proposed to this end, many working reasonably well in specific situations. However, no consensus has emerged on which of these methods is the best to use in practice. In part, this is due to the fact that testing their performance requires the generation of a comprehensive, standard set of synthetic benchmarks, a goal not yet fully achieved. Here, we present a type of benchmark that we call "closed", in which an initial network of known community structure is progressively converted into a second network whose communities are also known. This approach differs from all previously published ones, in which networks evolve toward randomness. The use of this type of benchmark allows us to monitor the transformation of the community structure of a network. Moreover, we can predict the optimal behavior of the variation of information, a measure of the quality of the partitions obtained, at any moment of the process. This enables us in many cases to determine the best partition among those suggested by different algorithms. Also, since any network can be used as a starting point, extensive studies and comparisons can be performed using a heterogeneous set of structures, including random ones. These properties make our benchmarks a general standard for comparing community detection algorithms.
[ { "version": "v1", "created": "Mon, 27 Feb 2012 12:29:47 GMT" }, { "version": "v2", "created": "Thu, 27 Dec 2012 10:52:26 GMT" } ]
2013-01-01T00:00:00
[ [ "Aldecoa", "Rodrigo", "" ], [ "Marín", "Ignacio", "" ] ]
TITLE: Closed benchmarks for network community structure characterization ABSTRACT: Characterizing the community structure of complex networks is a key challenge in many scientific fields. Very diverse algorithms and methods have been proposed to this end, many working reasonably well in specific situations. However, no consensus has emerged on which of these methods is the best to use in practice. In part, this is due to the fact that testing their performance requires the generation of a comprehensive, standard set of synthetic benchmarks, a goal not yet fully achieved. Here, we present a type of benchmark that we call "closed", in which an initial network of known community structure is progressively converted into a second network whose communities are also known. This approach differs from all previously published ones, in which networks evolve toward randomness. The use of this type of benchmark allows us to monitor the transformation of the community structure of a network. Moreover, we can predict the optimal behavior of the variation of information, a measure of the quality of the partitions obtained, at any moment of the process. This enables us in many cases to determine the best partition among those suggested by different algorithms. Also, since any network can be used as a starting point, extensive studies and comparisons can be performed using a heterogeneous set of structures, including random ones. These properties make our benchmarks a general standard for comparing community detection algorithms.
no_new_dataset
0.687827
1205.4463
Salah A. Aly
Salah A. Aly
Pilgrims Face Recognition Dataset -- HUFRD
5 pages, 13 images, 1 table of a new HUFRD work
null
null
null
cs.CV cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this work, we define a new pilgrims face recognition dataset, called HUFRD dataset. The new developed dataset presents various pilgrims' images taken from outside the Holy Masjid El-Harram in Makkah during the 2011-2012 Hajj and Umrah seasons. Such dataset will be used to test our developed facial recognition and detection algorithms, as well as assess in the missing and found recognition system \cite{crowdsensing}.
[ { "version": "v1", "created": "Sun, 20 May 2012 22:07:27 GMT" }, { "version": "v2", "created": "Sun, 30 Dec 2012 00:58:09 GMT" } ]
2013-01-01T00:00:00
[ [ "Aly", "Salah A.", "" ] ]
TITLE: Pilgrims Face Recognition Dataset -- HUFRD ABSTRACT: In this work, we define a new pilgrims face recognition dataset, called HUFRD dataset. The new developed dataset presents various pilgrims' images taken from outside the Holy Masjid El-Harram in Makkah during the 2011-2012 Hajj and Umrah seasons. Such dataset will be used to test our developed facial recognition and detection algorithms, as well as assess in the missing and found recognition system \cite{crowdsensing}.
new_dataset
0.714398
1212.6542
Dirk Beyer
Dirk Beyer and Stefan L\"owe
Explicit-Value Analysis Based on CEGAR and Interpolation
12 pages, 5 figures, 3 tables, 4 algorithms
null
null
MIP-1205
cs.SE cs.PL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Abstraction, counterexample-guided refinement, and interpolation are techniques that are essential to the success of predicate-based program analysis. These techniques have not yet been applied together to explicit-value program analysis. We present an approach that integrates abstraction and interpolation-based refinement into an explicit-value analysis, i.e., a program analysis that tracks explicit values for a specified set of variables (the precision). The algorithm uses an abstract reachability graph as central data structure and a path-sensitive dynamic approach for precision adjustment. We evaluate our algorithm on the benchmark set of the Competition on Software Verification 2012 (SV-COMP'12) to show that our new approach is highly competitive. In addition, we show that combining our new approach with an auxiliary predicate analysis scores significantly higher than the SV-COMP'12 winner.
[ { "version": "v1", "created": "Fri, 28 Dec 2012 17:47:10 GMT" } ]
2013-01-01T00:00:00
[ [ "Beyer", "Dirk", "" ], [ "Löwe", "Stefan", "" ] ]
TITLE: Explicit-Value Analysis Based on CEGAR and Interpolation ABSTRACT: Abstraction, counterexample-guided refinement, and interpolation are techniques that are essential to the success of predicate-based program analysis. These techniques have not yet been applied together to explicit-value program analysis. We present an approach that integrates abstraction and interpolation-based refinement into an explicit-value analysis, i.e., a program analysis that tracks explicit values for a specified set of variables (the precision). The algorithm uses an abstract reachability graph as central data structure and a path-sensitive dynamic approach for precision adjustment. We evaluate our algorithm on the benchmark set of the Competition on Software Verification 2012 (SV-COMP'12) to show that our new approach is highly competitive. In addition, we show that combining our new approach with an auxiliary predicate analysis scores significantly higher than the SV-COMP'12 winner.
no_new_dataset
0.707998
1212.6659
Raphael Pelossof
Raphael Pelossof and Zhiliang Ying
Focus of Attention for Linear Predictors
9 pages, 4 figures. arXiv admin note: substantial text overlap with arXiv:1105.0382
null
null
null
stat.ML cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a method to stop the evaluation of a prediction process when the result of the full evaluation is obvious. This trait is highly desirable in prediction tasks where a predictor evaluates all its features for every example in large datasets. We observe that some examples are easier to classify than others, a phenomenon which is characterized by the event when most of the features agree on the class of an example. By stopping the feature evaluation when encountering an easy- to-classify example, the predictor can achieve substantial gains in computation. Our method provides a natural attention mechanism for linear predictors where the predictor concentrates most of its computation on hard-to-classify examples and quickly discards easy-to-classify ones. By modifying a linear prediction algorithm such as an SVM or AdaBoost to include our attentive method we prove that the average number of features computed is O(sqrt(n log 1/sqrt(delta))) where n is the original number of features, and delta is the error rate incurred due to early stopping. We demonstrate the effectiveness of Attentive Prediction on MNIST, Real-sim, Gisette, and synthetic datasets.
[ { "version": "v1", "created": "Sat, 29 Dec 2012 20:23:48 GMT" } ]
2013-01-01T00:00:00
[ [ "Pelossof", "Raphael", "" ], [ "Ying", "Zhiliang", "" ] ]
TITLE: Focus of Attention for Linear Predictors ABSTRACT: We present a method to stop the evaluation of a prediction process when the result of the full evaluation is obvious. This trait is highly desirable in prediction tasks where a predictor evaluates all its features for every example in large datasets. We observe that some examples are easier to classify than others, a phenomenon which is characterized by the event when most of the features agree on the class of an example. By stopping the feature evaluation when encountering an easy- to-classify example, the predictor can achieve substantial gains in computation. Our method provides a natural attention mechanism for linear predictors where the predictor concentrates most of its computation on hard-to-classify examples and quickly discards easy-to-classify ones. By modifying a linear prediction algorithm such as an SVM or AdaBoost to include our attentive method we prove that the average number of features computed is O(sqrt(n log 1/sqrt(delta))) where n is the original number of features, and delta is the error rate incurred due to early stopping. We demonstrate the effectiveness of Attentive Prediction on MNIST, Real-sim, Gisette, and synthetic datasets.
no_new_dataset
0.710239
1212.5637
Claudio Gentile
Nicolo' Cesa-Bianchi, Claudio Gentile, Fabio Vitale, Giovanni Zappella
Random Spanning Trees and the Prediction of Weighted Graphs
Appeared in ICML 2010
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We investigate the problem of sequentially predicting the binary labels on the nodes of an arbitrary weighted graph. We show that, under a suitable parametrization of the problem, the optimal number of prediction mistakes can be characterized (up to logarithmic factors) by the cutsize of a random spanning tree of the graph. The cutsize is induced by the unknown adversarial labeling of the graph nodes. In deriving our characterization, we obtain a simple randomized algorithm achieving in expectation the optimal mistake bound on any polynomially connected weighted graph. Our algorithm draws a random spanning tree of the original graph and then predicts the nodes of this tree in constant expected amortized time and linear space. Experiments on real-world datasets show that our method compares well to both global (Perceptron) and local (label propagation) methods, while being generally faster in practice.
[ { "version": "v1", "created": "Fri, 21 Dec 2012 23:51:21 GMT" } ]
2012-12-27T00:00:00
[ [ "Cesa-Bianchi", "Nicolo'", "" ], [ "Gentile", "Claudio", "" ], [ "Vitale", "Fabio", "" ], [ "Zappella", "Giovanni", "" ] ]
TITLE: Random Spanning Trees and the Prediction of Weighted Graphs ABSTRACT: We investigate the problem of sequentially predicting the binary labels on the nodes of an arbitrary weighted graph. We show that, under a suitable parametrization of the problem, the optimal number of prediction mistakes can be characterized (up to logarithmic factors) by the cutsize of a random spanning tree of the graph. The cutsize is induced by the unknown adversarial labeling of the graph nodes. In deriving our characterization, we obtain a simple randomized algorithm achieving in expectation the optimal mistake bound on any polynomially connected weighted graph. Our algorithm draws a random spanning tree of the original graph and then predicts the nodes of this tree in constant expected amortized time and linear space. Experiments on real-world datasets show that our method compares well to both global (Perceptron) and local (label propagation) methods, while being generally faster in practice.
no_new_dataset
0.709265
1212.5701
Matthew Zeiler
Matthew D. Zeiler
ADADELTA: An Adaptive Learning Rate Method
6 pages
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent. The method requires no manual tuning of a learning rate and appears robust to noisy gradient information, different model architecture choices, various data modalities and selection of hyperparameters. We show promising results compared to other methods on the MNIST digit classification task using a single machine and on a large scale voice dataset in a distributed cluster environment.
[ { "version": "v1", "created": "Sat, 22 Dec 2012 15:46:49 GMT" } ]
2012-12-27T00:00:00
[ [ "Zeiler", "Matthew D.", "" ] ]
TITLE: ADADELTA: An Adaptive Learning Rate Method ABSTRACT: We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent. The method requires no manual tuning of a learning rate and appears robust to noisy gradient information, different model architecture choices, various data modalities and selection of hyperparameters. We show promising results compared to other methods on the MNIST digit classification task using a single machine and on a large scale voice dataset in a distributed cluster environment.
no_new_dataset
0.71022
1212.5766
Bach Ha
Nikhil R. Devanur, Bach Q. Ha, Jason D. Hartline
Prior-free Auctions for Budgeted Agents
null
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider prior-free auctions for revenue and welfare maximization when agents have a common budget. The abstract environments we consider are ones where there is a downward-closed and symmetric feasibility constraint on the probabilities of service of the agents. These environments include position auctions where slots with decreasing click-through rates are auctioned to advertisers. We generalize and characterize the envy-free benchmark from Hartline and Yan (2011) to settings with budgets and characterize the optimal envy-free outcomes for both welfare and revenue. We give prior-free mechanisms that approximate these benchmarks. A building block in our mechanism is a clinching auction for position auction environments. This auction is a generalization of the multi-unit clinching auction of Dobzinski et al. (2008) and a special case of the polyhedral clinching auction of Goel et al. (2012). For welfare maximization, we show that this clinching auction is a good approximation to the envy-free optimal welfare for position auction environments. For profit maximization, we generalize the random sampling profit extraction auction from Fiat et al. (2002) for digital goods to give a 10.0-approximation to the envy-free optimal revenue in symmetric, downward-closed environments. The profit maximization question is of interest even without budgets and our mechanism is a 7.5-approximation which improving on the 30.4 bound of Ha and Hartline (2012).
[ { "version": "v1", "created": "Sun, 23 Dec 2012 06:36:49 GMT" } ]
2012-12-27T00:00:00
[ [ "Devanur", "Nikhil R.", "" ], [ "Ha", "Bach Q.", "" ], [ "Hartline", "Jason D.", "" ] ]
TITLE: Prior-free Auctions for Budgeted Agents ABSTRACT: We consider prior-free auctions for revenue and welfare maximization when agents have a common budget. The abstract environments we consider are ones where there is a downward-closed and symmetric feasibility constraint on the probabilities of service of the agents. These environments include position auctions where slots with decreasing click-through rates are auctioned to advertisers. We generalize and characterize the envy-free benchmark from Hartline and Yan (2011) to settings with budgets and characterize the optimal envy-free outcomes for both welfare and revenue. We give prior-free mechanisms that approximate these benchmarks. A building block in our mechanism is a clinching auction for position auction environments. This auction is a generalization of the multi-unit clinching auction of Dobzinski et al. (2008) and a special case of the polyhedral clinching auction of Goel et al. (2012). For welfare maximization, we show that this clinching auction is a good approximation to the envy-free optimal welfare for position auction environments. For profit maximization, we generalize the random sampling profit extraction auction from Fiat et al. (2002) for digital goods to give a 10.0-approximation to the envy-free optimal revenue in symmetric, downward-closed environments. The profit maximization question is of interest even without budgets and our mechanism is a 7.5-approximation which improving on the 30.4 bound of Ha and Hartline (2012).
no_new_dataset
0.711481
1212.6246
Radford M. Neal
Chunyi Wang and Radford M. Neal
Gaussian Process Regression with Heteroscedastic or Non-Gaussian Residuals
null
null
null
null
stat.ML cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Gaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this paper, we propose a GP Regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for both modelsl. Experiments show that when the data is heteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and negative log-probability density) than standard GP regression. In addition, when the residual are Gaussian, our GPLC model is generally nearly as good as GPLV, while when the residuals are non-Gaussian, our GPLC model is better than GPLV.
[ { "version": "v1", "created": "Wed, 26 Dec 2012 20:45:48 GMT" } ]
2012-12-27T00:00:00
[ [ "Wang", "Chunyi", "" ], [ "Neal", "Radford M.", "" ] ]
TITLE: Gaussian Process Regression with Heteroscedastic or Non-Gaussian Residuals ABSTRACT: Gaussian Process (GP) regression models typically assume that residuals are Gaussian and have the same variance for all observations. However, applications with input-dependent noise (heteroscedastic residuals) frequently arise in practice, as do applications in which the residuals do not have a Gaussian distribution. In this paper, we propose a GP Regression model with a latent variable that serves as an additional unobserved covariate for the regression. This model (which we call GPLC) allows for heteroscedasticity since it allows the function to have a changing partial derivative with respect to this unobserved covariate. With a suitable covariance function, our GPLC model can handle (a) Gaussian residuals with input-dependent variance, or (b) non-Gaussian residuals with input-dependent variance, or (c) Gaussian residuals with constant variance. We compare our model, using synthetic datasets, with a model proposed by Goldberg, Williams and Bishop (1998), which we refer to as GPLV, which only deals with case (a), as well as a standard GP model which can handle only case (c). Markov Chain Monte Carlo methods are developed for both modelsl. Experiments show that when the data is heteroscedastic, both GPLC and GPLV give better results (smaller mean squared error and negative log-probability density) than standard GP regression. In addition, when the residual are Gaussian, our GPLC model is generally nearly as good as GPLV, while when the residuals are non-Gaussian, our GPLC model is better than GPLV.
no_new_dataset
0.712426
1105.2459
Rodrigo Aldecoa
Rodrigo Aldecoa and Ignacio Mar\'in
Deciphering Network Community Structure by Surprise
7 pages, 5 figures
PLoS ONE 6(9): e24195 (2011)
10.1371/journal.pone.0024195
null
q-bio.MN cs.SI physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The analysis of complex networks permeates all sciences, from biology to sociology. A fundamental, unsolved problem is how to characterize the community structure of a network. Here, using both standard and novel benchmarks, we show that maximization of a simple global parameter, which we call Surprise (S), leads to a very efficient characterization of the community structure of complex synthetic networks. Particularly, S qualitatively outperforms the most commonly used criterion to define communities, Newman and Girvan's modularity (Q). Applying S maximization to real networks often provides natural, well-supported partitions, but also sometimes counterintuitive solutions that expose the limitations of our previous knowledge. These results indicate that it is possible to define an effective global criterion for community structure and open new routes for the understanding of complex networks.
[ { "version": "v1", "created": "Thu, 12 May 2011 13:05:06 GMT" }, { "version": "v2", "created": "Fri, 5 Aug 2011 09:38:53 GMT" }, { "version": "v3", "created": "Fri, 21 Dec 2012 16:05:00 GMT" } ]
2012-12-24T00:00:00
[ [ "Aldecoa", "Rodrigo", "" ], [ "Marín", "Ignacio", "" ] ]
TITLE: Deciphering Network Community Structure by Surprise ABSTRACT: The analysis of complex networks permeates all sciences, from biology to sociology. A fundamental, unsolved problem is how to characterize the community structure of a network. Here, using both standard and novel benchmarks, we show that maximization of a simple global parameter, which we call Surprise (S), leads to a very efficient characterization of the community structure of complex synthetic networks. Particularly, S qualitatively outperforms the most commonly used criterion to define communities, Newman and Girvan's modularity (Q). Applying S maximization to real networks often provides natural, well-supported partitions, but also sometimes counterintuitive solutions that expose the limitations of our previous knowledge. These results indicate that it is possible to define an effective global criterion for community structure and open new routes for the understanding of complex networks.
no_new_dataset
0.709189
1211.3089
Yuheng Hu
Yuheng Hu, Ajita John, Fei Wang, Subbarao Kambhampati
ET-LDA: Joint Topic Modeling for Aligning Events and their Twitter Feedback
reference error, delete for now
null
null
null
cs.SI cs.AI cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
During broadcast events such as the Superbowl, the U.S. Presidential and Primary debates, etc., Twitter has become the de facto platform for crowds to share perspectives and commentaries about them. Given an event and an associated large-scale collection of tweets, there are two fundamental research problems that have been receiving increasing attention in recent years. One is to extract the topics covered by the event and the tweets; the other is to segment the event. So far these problems have been viewed separately and studied in isolation. In this work, we argue that these problems are in fact inter-dependent and should be addressed together. We develop a joint Bayesian model that performs topic modeling and event segmentation in one unified framework. We evaluate the proposed model both quantitatively and qualitatively on two large-scale tweet datasets associated with two events from different domains to show that it improves significantly over baseline models.
[ { "version": "v1", "created": "Tue, 13 Nov 2012 19:46:51 GMT" }, { "version": "v2", "created": "Fri, 21 Dec 2012 05:50:15 GMT" } ]
2012-12-24T00:00:00
[ [ "Hu", "Yuheng", "" ], [ "John", "Ajita", "" ], [ "Wang", "Fei", "" ], [ "Kambhampati", "Subbarao", "" ] ]
TITLE: ET-LDA: Joint Topic Modeling for Aligning Events and their Twitter Feedback ABSTRACT: During broadcast events such as the Superbowl, the U.S. Presidential and Primary debates, etc., Twitter has become the de facto platform for crowds to share perspectives and commentaries about them. Given an event and an associated large-scale collection of tweets, there are two fundamental research problems that have been receiving increasing attention in recent years. One is to extract the topics covered by the event and the tweets; the other is to segment the event. So far these problems have been viewed separately and studied in isolation. In this work, we argue that these problems are in fact inter-dependent and should be addressed together. We develop a joint Bayesian model that performs topic modeling and event segmentation in one unified framework. We evaluate the proposed model both quantitatively and qualitatively on two large-scale tweet datasets associated with two events from different domains to show that it improves significantly over baseline models.
no_new_dataset
0.709598
1212.5265
Tamal Ghosh Tamal Ghosh
Tamal Ghosh, Pranab K Dan
An Effective Machine-Part Grouping Algorithm to Construct Manufacturing Cells
null
Proceedings of Conference on Industrial Engineering (NCIE 2011)
null
null
cs.CE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The machine-part cell formation problem consists of creating machine cells and their corresponding part families with the objective of minimizing the inter-cell and intra-cell movement while maximizing the machine utilization. This article demonstrates a hybrid clustering approach for the cell formation problem in cellular manufacturing that conjoins Sorenson s similarity coefficient based method to form the production cells. Computational results are shown over the test datasets obtained from the past literature. The hybrid technique is shown to outperform the other methods proposed in literature and including powerful soft computing approaches such as genetic algorithms, genetic programming by exceeding the solution quality on the test problems.
[ { "version": "v1", "created": "Thu, 20 Dec 2012 15:51:13 GMT" } ]
2012-12-24T00:00:00
[ [ "Ghosh", "Tamal", "" ], [ "Dan", "Pranab K", "" ] ]
TITLE: An Effective Machine-Part Grouping Algorithm to Construct Manufacturing Cells ABSTRACT: The machine-part cell formation problem consists of creating machine cells and their corresponding part families with the objective of minimizing the inter-cell and intra-cell movement while maximizing the machine utilization. This article demonstrates a hybrid clustering approach for the cell formation problem in cellular manufacturing that conjoins Sorenson s similarity coefficient based method to form the production cells. Computational results are shown over the test datasets obtained from the past literature. The hybrid technique is shown to outperform the other methods proposed in literature and including powerful soft computing approaches such as genetic algorithms, genetic programming by exceeding the solution quality on the test problems.
no_new_dataset
0.710867
1212.5276
Marc Schoenauer
Mostepha Redouane Khouadjia (INRIA Saclay - Ile de France), Marc Schoenauer (INRIA Saclay - Ile de France, LRI), Vincent Vidal (DCSD), Johann Dr\'eo (TRT), Pierre Sav\'eant (TRT)
Multi-Objective AI Planning: Evaluating DAE-YAHSP on a Tunable Benchmark
7th International Conference on Evolutionary Multi-Criterion Optimization (2013) To appearr. arXiv admin note: text overlap with arXiv:0804.3965 by other authors
null
null
null
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
All standard AI planners to-date can only handle a single objective, and the only way for them to take into account multiple objectives is by aggregation of the objectives. Furthermore, and in deep contrast with the single objective case, there exists no benchmark problems on which to test the algorithms for multi-objective planning. Divide and Evolve (DAE) is an evolutionary planner that won the (single-objective) deterministic temporal satisficing track in the last International Planning Competition. Even though it uses intensively the classical (and hence single-objective) planner YAHSP, it is possible to turn DAE-YAHSP into a multi-objective evolutionary planner. A tunable benchmark suite for multi-objective planning is first proposed, and the performances of several variants of multi-objective DAE-YAHSP are compared on different instances of this benchmark, hopefully paving the road to further multi-objective competitions in AI planning.
[ { "version": "v1", "created": "Thu, 20 Dec 2012 21:26:17 GMT" } ]
2012-12-24T00:00:00
[ [ "Khouadjia", "Mostepha Redouane", "", "INRIA Saclay - Ile de France" ], [ "Schoenauer", "Marc", "", "INRIA Saclay - Ile de France, LRI" ], [ "Vidal", "Vincent", "", "DCSD" ], [ "Dréo", "Johann", "", "TRT" ], [ "Savéant", "Pierre", "", "TRT" ] ]
TITLE: Multi-Objective AI Planning: Evaluating DAE-YAHSP on a Tunable Benchmark ABSTRACT: All standard AI planners to-date can only handle a single objective, and the only way for them to take into account multiple objectives is by aggregation of the objectives. Furthermore, and in deep contrast with the single objective case, there exists no benchmark problems on which to test the algorithms for multi-objective planning. Divide and Evolve (DAE) is an evolutionary planner that won the (single-objective) deterministic temporal satisficing track in the last International Planning Competition. Even though it uses intensively the classical (and hence single-objective) planner YAHSP, it is possible to turn DAE-YAHSP into a multi-objective evolutionary planner. A tunable benchmark suite for multi-objective planning is first proposed, and the performances of several variants of multi-objective DAE-YAHSP are compared on different instances of this benchmark, hopefully paving the road to further multi-objective competitions in AI planning.
no_new_dataset
0.704783
1212.4458
Dan Burger
Dan Burger, Keivan G. Stassun, Joshua Pepper, Robert J. Siverd, Martin A. Paegert, Nathan M. De Lee
Filtergraph: A Flexible Web Application for Instant Data Visualization of Astronomy Datasets
4 pages, 1 figure. Originally presented at the ADASS XXII Conference in Champaign, IL on November 6, 2012. Published in the conference proceedings by ASP Conference Series (revised to include URL of web application)
null
null
null
astro-ph.IM cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Filtergraph is a web application being developed by the Vanderbilt Initiative in Data-intensive Astrophysics (VIDA) to flexibly handle a large variety of astronomy datasets. While current datasets at Vanderbilt are being used to search for eclipsing binaries and extrasolar planets, this system can be easily reconfigured for a wide variety of data sources. The user loads a flat-file dataset into Filtergraph which instantly generates an interactive data portal that can be easily shared with others. From this portal, the user can immediately generate scatter plots, histograms, and tables based on the dataset. Key features of the portal include the ability to filter the data in real time through user-specified criteria, the ability to select data by dragging on the screen, and the ability to perform arithmetic operations on the data in real time. The application is being optimized for speed in the context of very large datasets: for instance, plot generated from a stellar database of 3.1 million entries render in less than 2 seconds on a standard web server platform. This web application has been created using the Web2py web framework based on the Python programming language. Filtergraph is freely available at http://filtergraph.vanderbilt.edu/.
[ { "version": "v1", "created": "Tue, 18 Dec 2012 19:00:06 GMT" }, { "version": "v2", "created": "Wed, 19 Dec 2012 17:04:02 GMT" }, { "version": "v3", "created": "Thu, 20 Dec 2012 17:17:00 GMT" } ]
2012-12-21T00:00:00
[ [ "Burger", "Dan", "" ], [ "Stassun", "Keivan G.", "" ], [ "Pepper", "Joshua", "" ], [ "Siverd", "Robert J.", "" ], [ "Paegert", "Martin A.", "" ], [ "De Lee", "Nathan M.", "" ] ]
TITLE: Filtergraph: A Flexible Web Application for Instant Data Visualization of Astronomy Datasets ABSTRACT: Filtergraph is a web application being developed by the Vanderbilt Initiative in Data-intensive Astrophysics (VIDA) to flexibly handle a large variety of astronomy datasets. While current datasets at Vanderbilt are being used to search for eclipsing binaries and extrasolar planets, this system can be easily reconfigured for a wide variety of data sources. The user loads a flat-file dataset into Filtergraph which instantly generates an interactive data portal that can be easily shared with others. From this portal, the user can immediately generate scatter plots, histograms, and tables based on the dataset. Key features of the portal include the ability to filter the data in real time through user-specified criteria, the ability to select data by dragging on the screen, and the ability to perform arithmetic operations on the data in real time. The application is being optimized for speed in the context of very large datasets: for instance, plot generated from a stellar database of 3.1 million entries render in less than 2 seconds on a standard web server platform. This web application has been created using the Web2py web framework based on the Python programming language. Filtergraph is freely available at http://filtergraph.vanderbilt.edu/.
no_new_dataset
0.710377
1212.4871
Ramin Norousi
Ramin Norousi, Stephan Wickles, Christoph Leidig, Thomas Becker, Volker J. Schmid, Roland Beckmann, Achim Tresch
Automatic post-picking using MAPPOS improves particle image detection from Cryo-EM micrographs
null
null
null
null
stat.ML cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Cryo-electron microscopy (cryo-EM) studies using single particle reconstruction are extensively used to reveal structural information on macromolecular complexes. Aiming at the highest achievable resolution, state of the art electron microscopes automatically acquire thousands of high-quality micrographs. Particles are detected on and boxed out from each micrograph using fully- or semi-automated approaches. However, the obtained particles still require laborious manual post-picking classification, which is one major bottleneck for single particle analysis of large datasets. We introduce MAPPOS, a supervised post-picking strategy for the classification of boxed particle images, as additional strategy adding to the already efficient automated particle picking routines. MAPPOS employs machine learning techniques to train a robust classifier from a small number of characteristic image features. In order to accurately quantify the performance of MAPPOS we used simulated particle and non-particle images. In addition, we verified our method by applying it to an experimental cryo-EM dataset and comparing the results to the manual classification of the same dataset. Comparisons between MAPPOS and manual post-picking classification by several human experts demonstrated that merely a few hundred sample images are sufficient for MAPPOS to classify an entire dataset with a human-like performance. MAPPOS was shown to greatly accelerate the throughput of large datasets by reducing the manual workload by orders of magnitude while maintaining a reliable identification of non-particle images.
[ { "version": "v1", "created": "Wed, 19 Dec 2012 22:17:18 GMT" } ]
2012-12-21T00:00:00
[ [ "Norousi", "Ramin", "" ], [ "Wickles", "Stephan", "" ], [ "Leidig", "Christoph", "" ], [ "Becker", "Thomas", "" ], [ "Schmid", "Volker J.", "" ], [ "Beckmann", "Roland", "" ], [ "Tresch", "Achim", "" ] ]
TITLE: Automatic post-picking using MAPPOS improves particle image detection from Cryo-EM micrographs ABSTRACT: Cryo-electron microscopy (cryo-EM) studies using single particle reconstruction are extensively used to reveal structural information on macromolecular complexes. Aiming at the highest achievable resolution, state of the art electron microscopes automatically acquire thousands of high-quality micrographs. Particles are detected on and boxed out from each micrograph using fully- or semi-automated approaches. However, the obtained particles still require laborious manual post-picking classification, which is one major bottleneck for single particle analysis of large datasets. We introduce MAPPOS, a supervised post-picking strategy for the classification of boxed particle images, as additional strategy adding to the already efficient automated particle picking routines. MAPPOS employs machine learning techniques to train a robust classifier from a small number of characteristic image features. In order to accurately quantify the performance of MAPPOS we used simulated particle and non-particle images. In addition, we verified our method by applying it to an experimental cryo-EM dataset and comparing the results to the manual classification of the same dataset. Comparisons between MAPPOS and manual post-picking classification by several human experts demonstrated that merely a few hundred sample images are sufficient for MAPPOS to classify an entire dataset with a human-like performance. MAPPOS was shown to greatly accelerate the throughput of large datasets by reducing the manual workload by orders of magnitude while maintaining a reliable identification of non-particle images.
no_new_dataset
0.711017
1208.6119
An Zeng
Fuguo Zhang and An Zeng
Improving information filtering via network manipulation
6 pages, 5 figures
Europhysics Letters 100, 58005 (2012)
10.1209/0295-5075/100/58005
null
physics.soc-ph cs.SI physics.data-an
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recommender system is a very promising way to address the problem of overabundant information for online users. Though the information filtering for the online commercial systems received much attention recently, almost all of the previous works are dedicated to design new algorithms and consider the user-item bipartite networks as given and constant information. However, many problems for recommender systems such as the cold-start problem (i.e. low recommendation accuracy for the small degree items) are actually due to the limitation of the underlying user-item bipartite networks. In this letter, we propose a strategy to enhance the performance of the already existing recommendation algorithms by directly manipulating the user-item bipartite networks, namely adding some virtual connections to the networks. Numerical analyses on two benchmark data sets, MovieLens and Netflix, show that our method can remarkably improve the recommendation performance. Specifically, it not only improve the recommendations accuracy (especially for the small degree items), but also help the recommender systems generate more diverse and novel recommendations.
[ { "version": "v1", "created": "Thu, 30 Aug 2012 09:35:04 GMT" }, { "version": "v2", "created": "Fri, 31 Aug 2012 21:33:40 GMT" }, { "version": "v3", "created": "Wed, 21 Nov 2012 10:40:44 GMT" } ]
2012-12-20T00:00:00
[ [ "Zhang", "Fuguo", "" ], [ "Zeng", "An", "" ] ]
TITLE: Improving information filtering via network manipulation ABSTRACT: Recommender system is a very promising way to address the problem of overabundant information for online users. Though the information filtering for the online commercial systems received much attention recently, almost all of the previous works are dedicated to design new algorithms and consider the user-item bipartite networks as given and constant information. However, many problems for recommender systems such as the cold-start problem (i.e. low recommendation accuracy for the small degree items) are actually due to the limitation of the underlying user-item bipartite networks. In this letter, we propose a strategy to enhance the performance of the already existing recommendation algorithms by directly manipulating the user-item bipartite networks, namely adding some virtual connections to the networks. Numerical analyses on two benchmark data sets, MovieLens and Netflix, show that our method can remarkably improve the recommendation performance. Specifically, it not only improve the recommendations accuracy (especially for the small degree items), but also help the recommender systems generate more diverse and novel recommendations.
no_new_dataset
0.70978
1212.4788
Dominik Grimm dg
Dominik Grimm, Bastian Greshake, Stefan Kleeberger, Christoph Lippert, Oliver Stegle, Bernhard Sch\"olkopf, Detlef Weigel and Karsten Borgwardt
easyGWAS: An integrated interspecies platform for performing genome-wide association studies
null
null
null
null
q-bio.GN cs.CE cs.DL stat.AP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Motivation: The rapid growth in genome-wide association studies (GWAS) in plants and animals has brought about the need for a central resource that facilitates i) performing GWAS, ii) accessing data and results of other GWAS, and iii) enabling all users regardless of their background to exploit the latest statistical techniques without having to manage complex software and computing resources. Results: We present easyGWAS, a web platform that provides methods, tools and dynamic visualizations to perform and analyze GWAS. In addition, easyGWAS makes it simple to reproduce results of others, validate findings, and access larger sample sizes through merging of public datasets. Availability: Detailed method and data descriptions as well as tutorials are available in the supplementary materials. easyGWAS is available at http://easygwas.tuebingen.mpg.de/. Contact: dominik.grimm@tuebingen.mpg.de
[ { "version": "v1", "created": "Wed, 19 Dec 2012 18:39:06 GMT" } ]
2012-12-20T00:00:00
[ [ "Grimm", "Dominik", "" ], [ "Greshake", "Bastian", "" ], [ "Kleeberger", "Stefan", "" ], [ "Lippert", "Christoph", "" ], [ "Stegle", "Oliver", "" ], [ "Schölkopf", "Bernhard", "" ], [ "Weigel", "Detlef", "" ], [ "Borgwardt", "Karsten", "" ] ]
TITLE: easyGWAS: An integrated interspecies platform for performing genome-wide association studies ABSTRACT: Motivation: The rapid growth in genome-wide association studies (GWAS) in plants and animals has brought about the need for a central resource that facilitates i) performing GWAS, ii) accessing data and results of other GWAS, and iii) enabling all users regardless of their background to exploit the latest statistical techniques without having to manage complex software and computing resources. Results: We present easyGWAS, a web platform that provides methods, tools and dynamic visualizations to perform and analyze GWAS. In addition, easyGWAS makes it simple to reproduce results of others, validate findings, and access larger sample sizes through merging of public datasets. Availability: Detailed method and data descriptions as well as tutorials are available in the supplementary materials. easyGWAS is available at http://easygwas.tuebingen.mpg.de/. Contact: dominik.grimm@tuebingen.mpg.de
no_new_dataset
0.707398
1212.4347
Bonggun Shin
Bonggun Shin, Alice Oh
Bayesian Group Nonnegative Matrix Factorization for EEG Analysis
null
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose a generative model of a group EEG analysis, based on appropriate kernel assumptions on EEG data. We derive the variational inference update rule using various approximation techniques. The proposed model outperforms the current state-of-the-art algorithms in terms of common pattern extraction. The validity of the proposed model is tested on the BCI competition dataset.
[ { "version": "v1", "created": "Tue, 18 Dec 2012 13:35:38 GMT" } ]
2012-12-19T00:00:00
[ [ "Shin", "Bonggun", "" ], [ "Oh", "Alice", "" ] ]
TITLE: Bayesian Group Nonnegative Matrix Factorization for EEG Analysis ABSTRACT: We propose a generative model of a group EEG analysis, based on appropriate kernel assumptions on EEG data. We derive the variational inference update rule using various approximation techniques. The proposed model outperforms the current state-of-the-art algorithms in terms of common pattern extraction. The validity of the proposed model is tested on the BCI competition dataset.
no_new_dataset
0.710603
1212.3741
Qiqi Yan
Nikhil R. Devanur and Jason D. Hartline and Qiqi Yan
Envy Freedom and Prior-free Mechanism Design
null
null
null
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the provision of an abstract service to single-dimensional agents. Our model includes position auctions, single-minded combinatorial auctions, and constrained matching markets. When the agents' values are drawn from a distribution, the Bayesian optimal mechanism is given by Myerson (1981) as a virtual-surplus optimizer. We develop a framework for prior-free mechanism design and analysis. A good mechanism in our framework approximates the optimal mechanism for the distribution if there is a distribution; moreover, when there is no distribution this mechanism still performs well. We define and characterize optimal envy-free outcomes in symmetric single-dimensional environments. Our characterization mirrors Myerson's theory. Furthermore, unlike in mechanism design where there is no point-wise optimal mechanism, there is always a point-wise optimal envy-free outcome. Envy-free outcomes and incentive-compatible mechanisms are similar in structure and performance. We therefore use the optimal envy-free revenue as a benchmark for measuring the performance of a prior-free mechanism. A good mechanism is one that approximates the envy free benchmark on any profile of agent values. We show that good mechanisms exist, and in particular, a natural generalization of the random sampling auction of Goldberg et al. (2001) is a constant approximation.
[ { "version": "v1", "created": "Sun, 16 Dec 2012 00:06:41 GMT" } ]
2012-12-18T00:00:00
[ [ "Devanur", "Nikhil R.", "" ], [ "Hartline", "Jason D.", "" ], [ "Yan", "Qiqi", "" ] ]
TITLE: Envy Freedom and Prior-free Mechanism Design ABSTRACT: We consider the provision of an abstract service to single-dimensional agents. Our model includes position auctions, single-minded combinatorial auctions, and constrained matching markets. When the agents' values are drawn from a distribution, the Bayesian optimal mechanism is given by Myerson (1981) as a virtual-surplus optimizer. We develop a framework for prior-free mechanism design and analysis. A good mechanism in our framework approximates the optimal mechanism for the distribution if there is a distribution; moreover, when there is no distribution this mechanism still performs well. We define and characterize optimal envy-free outcomes in symmetric single-dimensional environments. Our characterization mirrors Myerson's theory. Furthermore, unlike in mechanism design where there is no point-wise optimal mechanism, there is always a point-wise optimal envy-free outcome. Envy-free outcomes and incentive-compatible mechanisms are similar in structure and performance. We therefore use the optimal envy-free revenue as a benchmark for measuring the performance of a prior-free mechanism. A good mechanism is one that approximates the envy free benchmark on any profile of agent values. We show that good mechanisms exist, and in particular, a natural generalization of the random sampling auction of Goldberg et al. (2001) is a constant approximation.
no_new_dataset
0.710829
1212.3964
Sourav Dutta
Suman K. Bera, Sourav Dutta, Ankur Narang and Souvik Bhattacherjee
Advanced Bloom Filter Based Algorithms for Efficient Approximate Data De-Duplication in Streams
41 pages
null
null
null
cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Applications involving telecommunication call data records, web pages, online transactions, medical records, stock markets, climate warning systems, etc., necessitate efficient management and processing of such massively exponential amount of data from diverse sources. De-duplication or Intelligent Compression in streaming scenarios for approximate identification and elimination of duplicates from such unbounded data stream is a greater challenge given the real-time nature of data arrival. Stable Bloom Filters (SBF) addresses this problem to a certain extent. . In this work, we present several novel algorithms for the problem of approximate detection of duplicates in data streams. We propose the Reservoir Sampling based Bloom Filter (RSBF) combining the working principle of reservoir sampling and Bloom Filters. We also present variants of the novel Biased Sampling based Bloom Filter (BSBF) based on biased sampling concepts. We also propose a randomized load balanced variant of the sampling Bloom Filter approach to efficiently tackle the duplicate detection. In this work, we thus provide a generic framework for de-duplication using Bloom Filters. Using detailed theoretical analysis we prove analytical bounds on the false positive rate, false negative rate and convergence rate of the proposed structures. We exhibit that our models clearly outperform the existing methods. We also demonstrate empirical analysis of the structures using real-world datasets (3 million records) and also with synthetic datasets (1 billion records) capturing various input distributions.
[ { "version": "v1", "created": "Mon, 17 Dec 2012 11:47:09 GMT" } ]
2012-12-18T00:00:00
[ [ "Bera", "Suman K.", "" ], [ "Dutta", "Sourav", "" ], [ "Narang", "Ankur", "" ], [ "Bhattacherjee", "Souvik", "" ] ]
TITLE: Advanced Bloom Filter Based Algorithms for Efficient Approximate Data De-Duplication in Streams ABSTRACT: Applications involving telecommunication call data records, web pages, online transactions, medical records, stock markets, climate warning systems, etc., necessitate efficient management and processing of such massively exponential amount of data from diverse sources. De-duplication or Intelligent Compression in streaming scenarios for approximate identification and elimination of duplicates from such unbounded data stream is a greater challenge given the real-time nature of data arrival. Stable Bloom Filters (SBF) addresses this problem to a certain extent. . In this work, we present several novel algorithms for the problem of approximate detection of duplicates in data streams. We propose the Reservoir Sampling based Bloom Filter (RSBF) combining the working principle of reservoir sampling and Bloom Filters. We also present variants of the novel Biased Sampling based Bloom Filter (BSBF) based on biased sampling concepts. We also propose a randomized load balanced variant of the sampling Bloom Filter approach to efficiently tackle the duplicate detection. In this work, we thus provide a generic framework for de-duplication using Bloom Filters. Using detailed theoretical analysis we prove analytical bounds on the false positive rate, false negative rate and convergence rate of the proposed structures. We exhibit that our models clearly outperform the existing methods. We also demonstrate empirical analysis of the structures using real-world datasets (3 million records) and also with synthetic datasets (1 billion records) capturing various input distributions.
no_new_dataset
0.710189
1212.3390
A Majumder
Anirban Majumder and Nisheeth Shrivastava
Know Your Personalization: Learning Topic level Personalization in Online Services
privacy, personalization
null
null
null
cs.LG cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Online service platforms (OSPs), such as search engines, news-websites, ad-providers, etc., serve highly pe rsonalized content to the user, based on the profile extracted from his history with the OSP. Although personalization (generally) leads to a better user experience, it also raises privacy concerns for the user---he does not know what is present in his profile and more importantly, what is being used to per sonalize content for him. In this paper, we capture OSP's personalization for an user in a new data structure called the person alization vector ($\eta$), which is a weighted vector over a set of topics, and present techniques to compute it for users of an OSP. Our approach treats OSPs as black-boxes, and extracts $\eta$ by mining only their output, specifical ly, the personalized (for an user) and vanilla (without any user information) contents served, and the differences in these content. We formulate a new model called Latent Topic Personalization (LTP) that captures the personalization vector into a learning framework and present efficient inference algorithms for it. We do extensive experiments for search result personalization using both data from real Google users and synthetic datasets. Our results show high accuracy (R-pre = 84%) of LTP in finding personalized topics. For Google data, our qualitative results show how LTP can also identifies evidences---queries for results on a topic with high $\eta$ value were re-ranked. Finally, we show how our approach can be used to build a new Privacy evaluation framework focused at end-user privacy on commercial OSPs.
[ { "version": "v1", "created": "Fri, 14 Dec 2012 04:12:21 GMT" } ]
2012-12-17T00:00:00
[ [ "Majumder", "Anirban", "" ], [ "Shrivastava", "Nisheeth", "" ] ]
TITLE: Know Your Personalization: Learning Topic level Personalization in Online Services ABSTRACT: Online service platforms (OSPs), such as search engines, news-websites, ad-providers, etc., serve highly pe rsonalized content to the user, based on the profile extracted from his history with the OSP. Although personalization (generally) leads to a better user experience, it also raises privacy concerns for the user---he does not know what is present in his profile and more importantly, what is being used to per sonalize content for him. In this paper, we capture OSP's personalization for an user in a new data structure called the person alization vector ($\eta$), which is a weighted vector over a set of topics, and present techniques to compute it for users of an OSP. Our approach treats OSPs as black-boxes, and extracts $\eta$ by mining only their output, specifical ly, the personalized (for an user) and vanilla (without any user information) contents served, and the differences in these content. We formulate a new model called Latent Topic Personalization (LTP) that captures the personalization vector into a learning framework and present efficient inference algorithms for it. We do extensive experiments for search result personalization using both data from real Google users and synthetic datasets. Our results show high accuracy (R-pre = 84%) of LTP in finding personalized topics. For Google data, our qualitative results show how LTP can also identifies evidences---queries for results on a topic with high $\eta$ value were re-ranked. Finally, we show how our approach can be used to build a new Privacy evaluation framework focused at end-user privacy on commercial OSPs.
no_new_dataset
0.71302
1212.3013
Gabriele Modena
K. Massoudi, G. Modena
Product/Brand extraction from WikiPedia
17 pages. Manuscript first creation date: November 27, 2009. At the time of first creation both authors were affiliated with the University of Amsterdam (The Netherlands)
null
null
null
cs.IR cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we describe the task of extracting product and brand pages from wikipedia. We present an experimental environment and setup built on top of a dataset of wikipedia pages we collected. We introduce a method for recognition of product pages modelled as a boolean probabilistic classification task. We show that this approach can lead to promising results and we discuss alternative approaches we considered.
[ { "version": "v1", "created": "Wed, 12 Dec 2012 23:25:46 GMT" } ]
2012-12-14T00:00:00
[ [ "Massoudi", "K.", "" ], [ "Modena", "G.", "" ] ]
TITLE: Product/Brand extraction from WikiPedia ABSTRACT: In this paper we describe the task of extracting product and brand pages from wikipedia. We present an experimental environment and setup built on top of a dataset of wikipedia pages we collected. We introduce a method for recognition of product pages modelled as a boolean probabilistic classification task. We show that this approach can lead to promising results and we discuss alternative approaches we considered.
no_new_dataset
0.709384
1212.2692
Ghazali Osman
Ghazali Osman, Muhammad Suzuri Hitam and Mohd Nasir Ismail
Enhanced skin colour classifier using RGB Ratio model
14 pages; International Journal on Soft Computing (IJSC) Vol.3, No.4, November 2012
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Skin colour detection is frequently been used for searching people, face detection, pornographic filtering and hand tracking. The presence of skin or non-skin in digital image can be determined by manipulating pixels colour or pixels texture. The main problem in skin colour detection is to represent the skin colour distribution model that is invariant or least sensitive to changes in illumination condition. Another problem comes from the fact that many objects in the real world may possess almost similar skin-tone colour such as wood, leather, skin-coloured clothing, hair and sand. Moreover, skin colour is different between races and can be different from a person to another, even with people of the same ethnicity. Finally, skin colour will appear a little different when different types of camera are used to capture the object or scene. The objective in this study is to develop a skin colour classifier based on pixel-based using RGB ratio model. The RGB ratio model is a newly proposed method that belongs under the category of an explicitly defined skin region model. This skin classifier was tested with SIdb dataset and two benchmark datasets; UChile and TDSD datasets to measure classifier performance. The performance of skin classifier was measured based on true positive (TF) and false positive (FP) indicator. This newly proposed model was compared with Kovac, Saleh and Swift models. The experimental results showed that the RGB ratio model outperformed all the other models in term of detection rate. The RGB ratio model is able to reduce FP detection that caused by reddish objects colour as well as be able to detect darkened skin and skin covered by shadow.
[ { "version": "v1", "created": "Wed, 12 Dec 2012 03:01:00 GMT" } ]
2012-12-13T00:00:00
[ [ "Osman", "Ghazali", "" ], [ "Hitam", "Muhammad Suzuri", "" ], [ "Ismail", "Mohd Nasir", "" ] ]
TITLE: Enhanced skin colour classifier using RGB Ratio model ABSTRACT: Skin colour detection is frequently been used for searching people, face detection, pornographic filtering and hand tracking. The presence of skin or non-skin in digital image can be determined by manipulating pixels colour or pixels texture. The main problem in skin colour detection is to represent the skin colour distribution model that is invariant or least sensitive to changes in illumination condition. Another problem comes from the fact that many objects in the real world may possess almost similar skin-tone colour such as wood, leather, skin-coloured clothing, hair and sand. Moreover, skin colour is different between races and can be different from a person to another, even with people of the same ethnicity. Finally, skin colour will appear a little different when different types of camera are used to capture the object or scene. The objective in this study is to develop a skin colour classifier based on pixel-based using RGB ratio model. The RGB ratio model is a newly proposed method that belongs under the category of an explicitly defined skin region model. This skin classifier was tested with SIdb dataset and two benchmark datasets; UChile and TDSD datasets to measure classifier performance. The performance of skin classifier was measured based on true positive (TF) and false positive (FP) indicator. This newly proposed model was compared with Kovac, Saleh and Swift models. The experimental results showed that the RGB ratio model outperformed all the other models in term of detection rate. The RGB ratio model is able to reduce FP detection that caused by reddish objects colour as well as be able to detect darkened skin and skin covered by shadow.
no_new_dataset
0.709818