text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
methods In approach kernel paper, this data. develop with an to exploiting manifold-valued we
In this paper, we develop an approach to exploiting kernel methods with manifold-valued data.
12,000
be representation. distinguished high-level ambiguity This can by
This ambiguity can be distinguished by high-level representation.
12,001
special of has considered. For case been processes, the Dirichlet streaming inference
For the special case of Dirichlet processes, streaming inference has been considered.
12,002
demonstrate text corpora. the the of efficacy in We documents algorithm on streaming clustering large,
We demonstrate the efficacy of the algorithm on clustering documents in large, streaming text corpora.
12,003
small. might be examples number when Therefore, is of training unreliable the they
Therefore, they might be unreliable when the number of training examples is small.
12,004
is robust overfitting to it small-size training Hence, data. on
Hence, it is robust to overfitting on small-size training data.
12,005
of asymmetric. is majority Unfortunately, models the graphical real-world
Unfortunately, the majority of real-world graphical models is asymmetric.
12,006
for the This relational case representations is when given. evidence even is
This is even the case for relational representations when evidence is given.
12,007
existing Markov such over-symmetric as First, networks. a logic relational all representation require approximations
First, all existing over-symmetric approximations require a relational representation such as Markov logic networks.
12,008
often computed the symmetries significantly, distribution highly the probabilities biased. making change Second, induced the
Second, the induced symmetries often change the distribution significantly, making the computed probabilities highly biased.
12,009
The while leads estimates probability therefore, unbiased. remaining framework, to improved
The framework, therefore, leads to improved probability estimates while remaining unbiased.
12,010
existing the approach Experiments demonstrate MCMC outperforms that algorithms.
Experiments demonstrate that the approach outperforms existing MCMC algorithms.
12,011
Analyzing corporations to Big Data im-prove help efficiency. can their
Analyzing Big Data can help corporations to im-prove their efficiency.
12,012
Semantic Machine-Learning proposea techniques HMC process, scalable We Rule-based reasoning. and also using
We also proposea Semantic HMC process, using scalable Machine-Learning techniques and Rule-based reasoning.
12,013
on We egocentric from task pose of estimation viewpoints. everyday focus hand the
We focus on the task of everyday hand pose estimation from egocentric viewpoints.
12,014
understanding limited. of them remains Nevertheless, our
Nevertheless, our understanding of them remains limited.
12,015
to contribute framework answer invert general this a question we To representations.
To answer this question we contribute a general framework to invert representations.
12,016
In compact learn is binary their embeddings manifolds to on work, intrinsic how this considered.
In this work, how to learn compact binary embeddings on their intrinsic manifolds is considered.
12,017
macromolecules of noisy structural central in is in variability projections biological a Cryo-EM. Classifying problem
Classifying structural variability in noisy projections of biological macromolecules is a central problem in Cryo-EM.
12,018
very easy and Finally, train are to conceptually and use. simple they
Finally, they are conceptually very simple and easy to train and use.
12,019
utility our of Lastly, the on through Internet experiments demonstrate collections. we formulations photo
Lastly, we demonstrate the utility of our formulations through experiments on Internet photo collections.
12,020
everyday objects. pose during interactions estimation we Importantly, with consider
Importantly, we consider pose estimation during everyday interactions with objects.
12,021
performance. work robust depth-based priors Past features for and are pose+viewpoint crucial shows that strong
Past work shows that strong pose+viewpoint priors and depth-based features are crucial for robust performance.
12,022
egocentric workspace. this call volume We an
We call this volume an egocentric workspace.
12,023
appearance property notable A workspace with location. is correlates hand that
A notable property is that hand appearance correlates with workspace location.
12,024
the This simplify performance. and improves architecture greatly
This greatly simplify the architecture and improves performance.
12,025
state-of-the-art method images performance pose hand RGB-D Our recognition egocentric provides from real-time. in
Our method provides state-of-the-art hand pose recognition performance from egocentric RGB-D images in real-time.
12,026
We architecture semantic purely for a introduce segmentation. feed-forward
We introduce a purely feed-forward architecture for semantic segmentation.
12,027
by classified a network. multilayer feedforward are Instead superpixels
Instead superpixels are classified by a feedforward multilayer network.
12,028
distortion a Results for of great type. its effectiveness variety demonstrates
Results demonstrates its effectiveness for a great variety of distortion type.
12,029
uses original other little the benefits about very only this Among image. information measure
Among other benefits this measure uses only very little information about the original image.
12,030
paper, we cost. accuracy time constrained the CNNs this of investigate In under
In this paper, we investigate the accuracy of CNNs under constrained time cost.
12,031
also This the in importance is designs. helpful of for network the factors understanding
This is also helpful for understanding the importance of the factors in network designs.
12,032
state-of-the-art our throughout. we of performance pipeline, stages Analysing show the
Analysing the stages of our pipeline, we show state-of-the-art performance throughout.
12,033
involved cost reduces runtime the in memory in learned turn CNNs. This and deploying the
This in turn reduces the memory and runtime cost involved in deploying the learned CNNs.
12,034
Besides, proposed on the results. more densities significant experiment tissues bring we approach various to
Besides, we experiment the proposed approach on various tissues densities to bring more significant results.
12,035
Base. explored from At challenging Data end, we breast medical some BIRADS this images
At this end, we explored some challenging breast images from BIRADS medical Data Base.
12,036
methods. promising with segmentation mass results the to first Our showed edges regard experimentations
Our first experimentations showed promising results with regard to the edges mass segmentation methods.
12,037
works This this in the discusses first area. paper achieved main
This paper discusses first the main works achieved in this area.
12,038
evaluated concluding paper. our and The results main are before showed
The main results are showed and evaluated before concluding our paper.
12,039
viewpoints. cases, of methods with as large Appearance-based appearances such change changes in fail
Appearance-based methods fail in such cases, as appearances change with large changes of viewpoints.
12,040
computed common different robust efficiently. This be can occlusions viewpoints, under measure to and is
This measure is robust to occlusions common under different viewpoints, and can be computed efficiently.
12,041
is retrieval cameras. stationary held demonstrated challenging Event using from and videos hand
Event retrieval is demonstrated using challenging videos from stationary and hand held cameras.
12,042
in detection image component fundamental and been detection Contour object segmentation has many a systems.
Contour detection has been a fundamental component in many image segmentation and object detection systems.
12,043
related and However, predicting mutually that claim we are recognizing objects two tasks. contours
However, we claim that recognizing objects and predicting contours are two mutually related tasks.
12,044
the This network scales of image different the to is applied four of section input.
This section of the network is applied to four different scales of the image input.
12,045
Why addressed complete? is we here Turing are why, question The
The question addressed here is why, Why are we Turing complete?
12,046
theory we theories. computing fusing build and a problem So by set
So we build a problem theory by fusing set and computing theories.
12,047
is conditions complete. Turing be to The the of last
The last of the conditions is to be Turing complete.
12,048
the most problems. answer would question to be: our to solve Then
Then the answer to our question would be: to solve most problems.
12,049
gradient descriptor, This efficient introduces spatiotemporal called local high histograms a (GBH). boundary paper
This paper introduces a high efficient local spatiotemporal descriptor, called gradient boundary histograms (GBH).
12,050
to GBH spatio-temporal on compute. is descriptor proposed are built gradients, fast which The simple
The proposed GBH descriptor is built on simple spatio-temporal gradients, which are fast to compute.
12,051
the of model tractable inference For marginal likelihood the approximations made. to be must
For tractable inference approximations to the marginal likelihood of the model must be made.
12,052
paper with idea this In compression. a we nested extend this variational
In this paper we extend this idea with a nested variational compression.
12,053
to Thereby, we able characteristic are the objects. properties determine of
Thereby, we are able to determine characteristic properties of the objects.
12,054
a a direct recognition comparison to similar rates. superior algorithm Moreover, shows
Moreover, a direct comparison to a similar algorithm shows superior recognition rates.
12,055
firing and evoked by learn can replay sequences stimuli. Neuronal circuits of patterns sensory
Neuronal circuits can learn and replay firing patterns evoked by sequences of sensory stimuli.
12,056
both a precise order sequences. propose of the timing We for learning event and mechanism
We propose a mechanism for learning both the order and precise timing of event sequences.
12,057
population Learned between another. populations synaptic to time one determine the necessary weights for activate
Learned synaptic weights between populations determine the time necessary for one population to activate another.
12,058
how of inherited The playback stochasticity is timings. determine process dynamics the sequence in learned
The dynamics of the playback process determine how stochasticity is inherited in learned sequence timings.
12,059
cooperation pressure Relatedness affect and selection the and on altruism. synergy
Relatedness and synergy affect the selection pressure on cooperation and altruism.
12,060
complementary ongoing contribute to we but ways. Here, this synthesis two in distinct
Here, we contribute to this ongoing synthesis in two distinct but complementary ways.
12,061
helping. on of This relatedness (or spatial qualitative broadens evolution of the the effects structure)
This broadens the qualitative effects of relatedness (or spatial structure) on the evolution of helping.
12,062
this exploit convolutional to features. propose masking we In via information shape paper,
In this paper, we propose to exploit shape information via masking convolutional features.
12,063
segments The treated (e.g., convolutional as super-pixels) on are proposal feature masks maps. the
The proposal segments (e.g., super-pixels) are treated as masks on the convolutional feature maps.
12,064
take will for emerge? long it to How them
How long will it take for them to emerge?
12,065
our firing-rate self-organization adaptation. questions these on based model address within We
We address these questions within our self-organization model based on firing-rate adaptation.
12,066
significant presents an to occlusion. humans there parsing This paper when is approach
This paper presents an approach to parsing humans when there is significant occlusion.
12,067
a We call parts. connected object flexible subtree composition of each
We call each connected subtree a flexible composition of object parts.
12,068
method a cues. occlusion novel for involves This learning
This involves a novel method for learning occlusion cues.
12,069
we a of inference search During over to models. flexible need mixture different
During inference we need to search over a mixture of different flexible models.
12,070
archaeologists. uninitiated put These off can aspects
These aspects can put off uninitiated archaeologists.
12,071
object spurred research methods. This recent improving in proposal
This spurred recent research in improving object proposal methods.
12,072
stable much Level iso-intensity ones. such more or are under especially the longer curves conditions,
Level or iso-intensity curves are much more stable under such conditions, especially the longer ones.
12,073
stable In corners on on iso-curves we them. this detect portions and long paper, identify
In this paper, we identify stable portions on long iso-curves and detect corners on them.
12,074
the signal. identification-verification It supervisory learned with is
It is learned with the identification-verification supervisory signal.
12,075
Markov the two-parameter overly Kimura complex or model). (the general model)
the Kimura two-parameter model) or overly complex (the general Markov model).
12,076
i.e. eyes salient can recognize on person based identities Human small regions,
Human eyes can recognize person identities based on small salient regions, i.e.
12,077
distinctive is reliable across human in pedestrian matching views. saliency and disjoint camera
human saliency is distinctive and reliable in pedestrian matching across disjoint camera views.
12,078
person methods outperforms approach on both Our state-of-the-art the re-identification datasets.
Our approach outperforms the state-of-the-art person re-identification methods on both datasets.
12,079
of unknown parameters. used compute These estimators are samples to the the Bayesian finally generated
These generated samples are finally used to compute the Bayesian estimators of the unknown parameters.
12,080
We a algorithm Problem. have concrete Traveling for Salesman created the
We have created a concrete algorithm for the Traveling Salesman Problem.
12,081
images We model of regions. generates descriptions their and a that language natural present
We present a model that generates natural language descriptions of images and their regions.
12,082
for importance of of classification. the parts investigate action and the tasks We attribute
We investigate the importance of parts for the tasks of action and attribute classification.
12,083
biomolecules important most Proteins for organisms. the are living
Proteins are the most important biomolecules for living organisms.
12,084
are characterization, for utilized protein classification. identification MTFs and
MTFs are utilized for protein characterization, identification and classification.
12,085
track proposed of origin geometric of the method protein topological slicing The to invariants. is
The method of slicing is proposed to track the geometric origin of protein topological invariants.
12,086
and all-atom coarse-grained representations of MTFs are constructed. Both
Both all-atom and coarse-grained representations of MTFs are constructed.
12,087
end, filtration based correlation developed. a this To is matrix
To this end, a correlation matrix based filtration is developed.
12,088
excellent simulation and between homology prediction An molecular consistence is persistent found. our dynamics
An excellent consistence between our persistent homology prediction and molecular dynamics simulation is found.
12,089
topology-function the This reveals work of proteins. relationship
This work reveals the topology-function relationship of proteins.
12,090
and essentially qualitative limited data (CIA). a such is analysis identification characterization, However, to success
However, such a success is essentially limited to qualitative data characterization, identification and analysis (CIA).
12,091
a we objective-oriented protocol homology general outline methods. this persistent work, to construct In
In this work, we outline a general protocol to construct objective-oriented persistent homology methods.
12,092
EGG generator waveform assembly those neuron based modeled gastric data. using on morphological
EGG waveform generator based on gastric morphological neuron assembly modeled using those data.
12,093
correlation to our based further filtration introduced findings. matrix verify A is
A correlation matrix based filtration is introduced to further verify our findings.
12,094
therapy lumbar as injection images These guidance ultrasound for of act treatment for radiculopathy.
These ultrasound images act as guidance for injection therapy for treatment of lumbar radiculopathy.
12,095
resorting protein the protein one predict without can to As flexibility Hamiltonian. such, interaction
As such, one can predict protein flexibility without resorting to the protein interaction Hamiltonian.
12,096
This algorithm fast work analysis for macromolecules. (fFRI) the a introduces of large FRI flexibility
This work introduces a fast FRI (fFRI) algorithm for the flexibility analysis of large macromolecules.
12,097
The proposed the complexity O(N). further fFRI to reduces computational
The proposed fFRI further reduces the computational complexity to O(N).
12,098
for protein analysis (aFRI) Additionally, the FRI of algorithms anisotropic we propose dynamics. collective
Additionally, we propose anisotropic FRI (aFRI) algorithms for the analysis of protein collective dynamics.
12,099