text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
in validate $\delta$-medoids namely empirically We music and motion domains, analysis. analysis two
We empirically validate $\delta$-medoids in two domains, namely music analysis and motion analysis.
13,000
regularization of surface functional. area used function The as was the
The function of surface area was used as the regularization functional.
13,001
dynamic because estimate previous variations. image current regularization on character depends has and The
The regularization has dynamic character because depends on current and previous image estimate variations.
13,002
the distribution Monte Currently Carlo probability used coincidences. of methods approximate
Currently used Monte Carlo methods approximate the probability distribution of coincidences.
13,003
an is LSTM for recognition. ideal candidate text Deep
Deep LSTM is an ideal candidate for text recognition.
13,004
becomes difficult range intractable. segmentation, context long learning very is and computationally Without
Without segmentation, learning very long range context is difficult and becomes computationally intractable.
13,005
decisions alternative Therefore, are soft at the level. pre-processing needed
Therefore, alternative soft decisions are needed at the pre-processing level.
13,006
filters errors. verification language outputs segmentation framework models using eliminates wrong Thus recognition and the
Thus the verification framework using language models eliminates wrong segmentation outputs and filters recognition errors.
13,007
sequential standard Resampling is generally Monte Carlo in filters step methods. a more particle and
Resampling is a standard step in particle filters and more generally sequential Monte Carlo methods.
13,008
particles. resampling We for algorithm, called chopthin, an present weighted
We present an algorithm, called chopthin, for resampling weighted particles.
13,009
resampling algorithm outperforms studies standard Simulation chopthin the consistently that show methods.
Simulation studies show that the chopthin algorithm consistently outperforms standard resampling methods.
13,010
implicitly a bound size. the sample guarantees effective lower on It
It implicitly guarantees a lower bound on the effective sample size.
13,011
making The can useful. implemented it algorithm practically efficiently, be
The algorithm can be implemented efficiently, making it practically useful.
13,012
expected the of show the effort We particles. linear in that computational number is
We show that the expected computational effort is linear in the number of particles.
13,013
available. CRAN), R for (on Python Matlab Implementations C++, are and
Implementations for C++, R (on CRAN), Python and Matlab are available.
13,014
article, of estimation we problem images. this tackle monocular depth the from In single
In this article, we tackle the problem of depth estimation from single monocular images.
13,015
are common field processing life. Image every our in applications daily of
Image processing applications are common in every field of our daily life.
13,016
Therefore, solutions processing. hardware heterogeneous on are application-specialized image for based used architectures
Therefore, application-specialized hardware solutions based on heterogeneous architectures are used for image processing.
13,017
scratch which and error-prone systems processing inflexible. time-consuming, Nowadays, image from started is are
Nowadays, image processing systems are started from scratch which is time-consuming, error-prone and inflexible.
13,018
This systems. introduces paper top new approach image down a for holistic processing
This paper introduces a new holistic top down approach for image processing systems.
13,019
architecture. to want and close Second, between system abstract we the gap system the the
Second, we want to close the gap between the abstract system and the system architecture.
13,020
Information data with and research field such are visualization two mining goal.
Information visualization and data mining are two research field with such goal.
13,021
deep improved trained the considerably performance Recently, object from recognition. have CNNs of large datasets
Recently, deep CNNs trained from large datasets have considerably improved the performance of object recognition.
13,022
using non-uniform image motion prior. a is removed patch-level blur Finally, model by deblurring
Finally, motion blur is removed by a non-uniform deblurring model using patch-level image prior.
13,023
or computer-generated with The results drift only are surrogate containing compared only time series microsaccades.
The results are compared with computer-generated surrogate time series containing only drift or only microsaccades.
13,024
analyses components drift. to between due Our show a binocular dependence FEM predominantly velocity
Our analyses show a binocular dependence between FEM velocity components predominantly due to drift.
13,025
SVMs Ensemble calibrating Exemplar the We method of for present a model.
We present a method for calibrating the Ensemble of Exemplar SVMs model.
13,026
with We on descriptors. trained experiment state-of-the-art CNN EE-SVM
We experiment with EE-SVM trained on state-of-the-art CNN descriptors.
13,027
normalized algorithm. fast this a first purpose, For we develop cuts
For this purpose, we first develop a fast normalized cuts algorithm.
13,028
effective propose a that high-performance multiscale information. makes then of use segmenter hierarchical We
We then propose a high-performance hierarchical segmenter that makes effective use of multiscale information.
13,029
their animals on Humans sensory noisy and input. base other decisions
Humans and other animals base their decisions on noisy sensory input.
13,030
been the computations that work such Much decisions. devoted understanding underly has to therefore
Much work has therefore been devoted to understanding the computations that underly such decisions.
13,031
understood. is impact well the judgments of However, in correlations on sensory perceptual noise not
However, the impact of correlations in sensory noise on perceptual judgments is not well understood.
13,032
impact correlations Here stimulus decision we noise in examine correlations sensory together how with making.
Here we examine how stimulus correlations together with correlations in sensory noise impact decision making.
13,033
distractors are of correlated. the and We measurements assume both the the that observer's stimuli
We assume that both the distractors and the observer's measurements of the stimuli are correlated.
13,034
impact structured in strongly therefore can to responses perceptual Correlations judgments. stimuli neural
Correlations in neural responses to structured stimuli can therefore strongly impact perceptual judgments.
13,035
$T$ for means unique that Here phylogeny is perfect characters. these the `capture'
Here `capture' means that $T$ is the unique perfect phylogeny for these characters.
13,036
combinatorial of method. Our this proof on probabilistic result based is the short
Our short proof of this combinatorial result is based on the probabilistic method.
13,037
a category computer localization Object problem is vision. in challenging
Object category localization is a challenging problem in computer vision.
13,038
annotations object supervised requires box instances. training bounding Standard of
Standard supervised training requires bounding box annotations of object instances.
13,039
is annotation time-consuming in learning. process sidestepped weakly supervised This
This time-consuming annotation process is sidestepped in weakly supervised learning.
13,040
information. and In layer adopt appearance feature motion information which concatenation combines we a addition,
In addition, we adopt a feature concatenation layer which combines appearance information and motion information.
13,041
an the for propose problem. colony We ant optimization algorithm
We propose an ant colony optimization algorithm for the problem.
13,042
role plays industry. an research and important in HFR both biometrics
HFR plays an important role in both biometrics research and industry.
13,043
degree overestimation when significant of spatial marked. heterogeneity The becomes is
The degree of overestimation becomes significant when spatial heterogeneity is marked.
13,044
. on those examples accelerating illustrate a of We series features
We illustrate those accelerating features on a series of examples .
13,045
and techniques, fast memory to efficient. is ConF accurate, Compared nearest-neighbour more standard
Compared to standard nearest-neighbour techniques, ConF is more accurate, fast and memory efficient.
13,046
discarding inaccurate frequencies, clonal etc. computation specific genotypes clone of
inaccurate computation of clonal frequencies, discarding clone specific genotypes etc.
13,047
tested also our We simulated different on model data.
We also tested our model on different simulated data.
13,048
heterogeneity. methods that results shows HetFHMM outperforms tumor infer recent which The
The results shows that HetFHMM outperforms recent methods which infer tumor heterogeneity.
13,049
research novel area. heterogeneity tumor Therefore, approach is in HetFHMM a
Therefore, HetFHMM is a novel approach in tumor heterogeneity research area.
13,050
This around ability sustained distracting of addition object. dots be with the can
This ability can be sustained with addition of distracting dots around the object.
13,051
has capability such been algorithmically computers. reproduced on However, not
However, such capability has not been reproduced algorithmically on computers.
13,052
into new a set subsets. algorithm We non-disjoint introduce that multiple of groups dots a
We introduce a new algorithm that groups a set of dots into multiple non-disjoint subsets.
13,053
a spanning into It the cue. dots connects the using tree proximity
It connects the dots into a spanning tree using the proximity cue.
13,054
is O($n$). of number generated polygons The
The number of polygons generated is O($n$).
13,055
both these effectiveness show algorithms. and We of to images synthetic used natural
We used both natural and synthetic images to show effectiveness of these algorithms.
13,056
considerations. introduce normative a formal distinction between we statistical Further, and
Further, we introduce a formal distinction between statistical and normative considerations.
13,057
single-target tracker paper problem addresses This performance the evaluation. of
This paper addresses the problem of single-target tracker performance evaluation.
13,058
statistical equivalence ranking-based The significance and in methodology tracker practical terms differences. of addresses
The ranking-based methodology addresses tracker equivalence in terms of statistical significance and practical differences.
13,059
introduced. annotations dataset several visual A with with fully-annotated is per-frame attributes
A fully-annotated dataset with per-frame annotations with several visual attributes is introduced.
13,060
the This to sophistically most dataset makes annotated constructed date. and it
This makes it the most sophistically constructed and annotated dataset to date.
13,061
as presented of third-party easy system trackers is allowing multi-platform evaluation A well. integration
A multi-platform evaluation system allowing easy integration of third-party trackers is presented as well.
13,062
difficulty the is of perspective An of analysis dataset the from exhaustive carried out. tracking
An exhaustive analysis of the dataset from the perspective of tracking difficulty is carried out.
13,063
new comparison a To facilitate proposed. technique is tracker performance visualization
To facilitate tracker comparison a new performance visualization technique is proposed.
13,064
Discrete-time a of approach for capturing system. models are dynamics population traditional host-parasitoid the
Discrete-time models are the traditional approach for capturing population dynamics of a host-parasitoid system.
13,065
and debate. diversification have Sino-Tibetan populations been The hot a of long-standing origin
The origin and diversification of Sino-Tibetan populations have been a long-standing hot debate.
13,066
However, Tibetan of the populations information genetic topic from this clear. limited far keeps
However, the limited genetic information of Tibetan populations keeps this topic far from clear.
13,067
identification. for also dataset valuable are generated The here forensic
The dataset generated here are also valuable for forensic identification.
13,068
the degraded when However, dealing images. performance color PCANet be with may of
However, the performance of PCANet may be degraded when dealing with color images.
13,069
method CEoptim optimization. the the implementation package R for new The CE of provides
The new package CEoptim provides the R implementation of the CE method for optimization.
13,070
We some describe optimization useful for CE general as the and modifications. well methodology
We describe the general CE methodology for optimization and well as some useful modifications.
13,071
compromising orders show results two of without of performance. complexity reduction Simulation a unmixing magnitude
Simulation results show a complexity reduction of two orders of magnitude without compromising unmixing performance.
13,072
is wide-spread Bistability bacteria considered e.g. cells, and eukaryotic among useful
Bistability is considered wide-spread among bacteria and eukaryotic cells, useful e.g.
13,073
for switching. induction, epigenetic and bet hedging, enzyme
for enzyme induction, bet hedging, and epigenetic switching.
13,074
However, deterministic well-mixed has described with this stochastic or been dynamic models. mostly phenomenon
However, this phenomenon has mostly been described with deterministic dynamic or well-mixed stochastic models.
13,075
withdrawn the by authour. This been has paper
This paper has been withdrawn by the authour.
13,076
on synthetic real-world well results as data. encouraging as Experiments show
Experiments show encouraging results on synthetic as well as real-world data.
13,077
cut proposes of This algorithm for a clustering. variant paper the spectral normalized
This paper proposes a variant of the normalized cut algorithm for spectral clustering.
13,078
to performance the of is presented the empirical study examine algorithm. An
An empirical study is presented to examine the performance of the algorithm.
13,079
of data machine applications occur categorical learning. many Multivariate in
Multivariate categorical data occur in many applications of machine learning.
13,080
variables is difficulties of of vectors categorical One sparsity. these main with the
One of the main difficulties with these vectors of categorical variables is sparsity.
13,081
data. have significant gained improvement in tasks supervised models with Recent this
Recent models have gained significant improvement in supervised tasks with this data.
13,082
models between continuous them. to observations a similarities embed in capture These space
These models embed observations in a continuous space to capture similarities between them.
13,083
sampling for variational based inference We recent in developments based derive our model on inference.
We derive inference for our model based on recent developments in sampling based variational inference.
13,084
storage in stored that are table All hash possible minimum space. a requires transitions state
All possible state transitions are stored in a hash table that requires minimum storage space.
13,085
The each is individual the of predicted of probability. the confidence player's product team's formation
The confidence of the predicted team's formation is the product of each individual player's probability.
13,086
is low-level framework visual and on high-level hand-crafted the basis multiple Our features. built of
Our framework is built on the basis of multiple low-level hand-crafted and high-level visual features.
13,087
is proposals basic training and iterate generating idea networks. The to between automatically region convolutional
The basic idea is to iterate between automatically generating region proposals and training convolutional networks.
13,088
These vise masks the for improving gradually and versa. segmentation recover steps two networks,
These two steps gradually recover segmentation masks for improving the networks, and vise versa.
13,089
problems optimization. Many formulated high sparse nonconvex are as dimensional learning
Many high dimensional sparse learning problems are formulated as nonconvex optimization.
13,090
the limits relaxations. In paper, this of we statistical study convex
In this paper, we study the statistical limits of convex relaxations.
13,091
illustrated Bayesian is regression spline by example. performance Its free-knot a
Its performance is illustrated by a Bayesian free-knot spline regression example.
13,092
unordered experiment sequentially this and on both Internet We captured method images. images
We experiment this method on both sequentially captured images and unordered Internet images.
13,093
and robustness, efficiency. in accuracy, The demonstrate experiments its strength
The experiments demonstrate its strength in robustness, accuracy, and efficiency.
13,094
systems recognition increasing Datasets steadily object for training size. in are
Datasets for training object recognition systems are steadily increasing in size.
13,095
the We discriminatively features. focus templates oriented defined on gradient paradigm popular trained of on
We focus on the popular paradigm of discriminatively trained templates defined on oriented gradient features.
13,096
Most think works. of basic us about we some how know facts time
Most of us think we know some basic facts about how time works.
13,097
excellent of These our ordered understanding are events. ones intuitions for conception conscious
These intuitions are excellent ones for understanding our conscious conception of ordered events.
13,098
motor proteins most transported In cells, of by along microtubule. cargos are
In cells, most of cargos are transported by motor proteins along microtubule.
13,099