text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
we problem properties generator. rethink make this paper, of In the a proposal good what
In this paper, we rethink the problem of what properties make a good proposal generator.
35,100
an social human-computer interaction. non-verbal processing and behavior Gaze in important is cue signal
Gaze behavior is an important non-verbal cue in social signal processing and human-computer interaction.
35,101
in computations. The is executing extremely data-extensive challenge their biggest DNNs
The biggest challenge in executing DNNs is their extremely data-extensive computations.
35,102
data the deployment. directly networks quantizing %However, accuracy deep is in by limited system
%However, the system accuracy is limited by quantizing data directly in deep networks deployment.
35,103
Previous inter-layer mainly works signals neglected. discretize on mainly weights while are focus
Previous works mainly focus on weights discretize while inter-layer signals are mainly neglected.
35,104
example. the DNNs system implement memristor-based a proposed on the SNC We deployment as
We implement the proposed DNNs on the memristor-based SNC system as a deployment example.
35,105
of finite previously a neurons. for uniform We size effects globally developed theory coupled generalize
We generalize a previously developed theory of finite size effects for uniform globally coupled neurons.
35,106
discriminative an are and around object? views where But informative such
But where are such informative and discriminative views around an object?
35,107
and venues. this work, travel are as destination location In taken business
In this work, travel destination and business location are taken as venues.
35,108
by is applications. very a Discovering important for a photo venue context-aware
Discovering a venue by a photo is very important for context-aware applications.
35,109
data. heterogeneous discovery goal venue multimodal is Our fine-grained from social
Our goal is fine-grained venue discovery from heterogeneous social multimodal data.
35,110
on proposed the of confirm method. results feasibility the this Experimental dataset
Experimental results on this dataset confirm the feasibility of the proposed method.
35,111
estimates likelihood marginal unbiased. is An additional benefit are that
An additional benefit is that marginal likelihood estimates are unbiased.
35,112
AU. The of determined the Gaussian is size the of the and by amplitude intensity
The amplitude and size of the Gaussian is determined by the intensity of the AU.
35,113
has bandstop A in proposed varactor-based been article. tunable filter this
A varactor-based tunable bandstop filter has been proposed in this article.
35,114
have networks classification. neural used Convolutional for widely image (CNNs) been
Convolutional neural networks (CNNs) have been widely used for image classification.
35,115
propose novel we a improve convolutional the To learning prototype (CPL). learning robustness, framework called
To improve the robustness, we propose a novel learning framework called convolutional prototype learning (CPL).
35,116
train classification we to criteria of CPL, framework the design the Under multiple network.
Under the framework of CPL, we design multiple classification criteria to train the network.
35,117
Positive geometry matrices covariance The of of (SPD) that space Definite matrices. the Symmetric is
The space geometry of the covariance matrices is that of Symmetric Positive Definite (SPD) matrices.
35,118
systems Automated rely vehicles. proper passing on classification toll the of
Automated toll systems rely on proper classification of the passing vehicles.
35,119
vehicle. whole obtain about To the information
To obtain information about the whole vehicle.
35,120
Existing interactive as promising approaches venue. a propose search image
Existing approaches propose interactive image search as a promising venue.
35,121
three and datasets baselines three experimental We extensive on outperform settings.
We outperform three baselines on three datasets and extensive experimental settings.
35,122
this interpretation. with comes of approach complications However,
However, this approach comes with complications of interpretation.
35,123
Optimal parameter Instrumental Linear The approximation the (OLIVA). in called Variables the Approximation is linear
The parameter in the linear approximation is called the Optimal Linear Instrumental Variables Approximation (OLIVA).
35,124
instrument is in unknown the be identified. The estimand not may IV and
The instrument in the IV estimand is unknown and may not be identified.
35,125
discuss least We also to weighted criteria. extensions squares
We also discuss extensions to weighted least squares criteria.
35,126
excellent suggest performance an simulations finite sample the Carlo for proposed inferences. Monte
Monte Carlo simulations suggest an excellent finite sample performance for the proposed inferences.
35,127
set a is At training-time, filters. explicit the boundary with provided of network separate
At training-time, the network is provided with a separate set of explicit boundary filters.
35,128
tracking in a compared plays role performance trackers. Correlation major filter existing improved to
Correlation filter plays a major role in improved tracking performance compared to existing trackers.
35,129
The the tracker of the to location predict correlation the response target. uses adaptive
The tracker uses the adaptive correlation response to predict the location of the target.
35,130
frame were proposed accuracy high correlation recently varieties of and trackers with rates. Many
Many varieties of correlation trackers were proposed recently with high accuracy and frame rates.
35,131
approach validated sequences. Benchmark Tracking Object is The using
The approach is validated using Object Tracking Benchmark sequences.
35,132
The optimize frame-wise usually likelihood the employ maximal existing a models. to loss methods
The existing methods usually employ a frame-wise maximal likelihood loss to optimize the models.
35,133
can that performance. recognition text substantially boost show scene Experimental EP results the
Experimental results show that the EP can substantially boost scene text recognition performance.
35,134
constraint specific search requires good problems of often heuristics. or Effective choosing solving
Effective solving of constraint problems often requires choosing good or specific search heuristics.
35,135
is to heuristics give search robust which automatic The goal performance. obtain
The goal is to obtain automatic search heuristics which give robust performance.
35,136
our adaptive than technique experiments is original heuristics. more show that the robust Preliminary search
Preliminary experiments show that our adaptive technique is more robust than the original search heuristics.
35,137
outperform the also heuristics. can original It
It can also outperform the original heuristics.
35,138
Using reduction, analysis. significant we PWM-based dimension identify subspaces for
Using dimension reduction, we identify significant PWM-based subspaces for analysis.
35,139
ML-based sequence PWM-based These subspaces tool. an form analysis
These PWM-based subspaces form an ML-based sequence analysis tool.
35,140
improved the TFs. on performance mammalian It also
It also improved the performance on mammalian TFs.
35,141
easily to tools PWM-based future well information. as is more as The include extendable ensemble
The ensemble is easily extendable to include more tools as well as future PWM-based information.
35,142
personalized full Reaching the on precision of genome of interpretation. the medicine potential depends quality
Reaching the full potential of precision medicine depends on the quality of personalized genome interpretation.
35,143
the compared better the original that achieve results results show proposed with Evaluation FlowNet. network
Evaluation results show that the proposed network achieve better results compared with the original FlowNet.
35,144
This block-matching optical flow called proposes (ABMOF). algorithm event-driven adaptive OF an paper
This paper proposes an event-driven OF algorithm called adaptive block-matching optical flow (ABMOF).
35,145
ABMOF accumulated events. time of uses slices DVS
ABMOF uses time slices of accumulated DVS events.
35,146
results. The OF based time slices adaptively input events are on rotated the and
The time slices are adaptively rotated based on the input events and OF results.
35,147
conventional standards as Lucas-Kanade Results such achieves accuracy (LK). comparable show that to ABMOF
Results show that ABMOF achieves comparable accuracy to conventional standards such as Lucas-Kanade (LK).
35,148
implemented. our adapted using also is An slices LK method
An LK method using our adapted slices is also implemented.
35,149
strategies influence and this sampling conformational how explore space how might predictions? do Namely,
Namely, how do sampling strategies explore conformational space and how might this influence predictions?
35,150
we kinetics. compare in describing thermodynamics MSM Additionally, and estimators
Additionally, we compare MSM estimators in describing thermodynamics and kinetics.
35,151
dilated equipped we convolutional blocks a different generic classification of Then, with rates. network design
Then, we design a generic classification network equipped with convolutional blocks of different dilated rates.
35,152
obtains apparent performance approach Despite state-of-the-arts. over proposed the our superior simplicity,
Despite the apparent simplicity, our proposed approach obtains superior performance over state-of-the-arts.
35,153
of different facial landmarks. distortions into expressions categories facial requires capturing regional Classifying
Classifying facial expressions into different categories requires capturing regional distortions of facial landmarks.
35,154
best are of results results aware these are Both we the of.
Both of these results are the best results we are aware of.
35,155
considered. for adaptive is worst-case problem robust The beamforming signal general-rank model
The worst-case robust adaptive beamforming problem for general-rank signal model is considered.
35,156
algorithm to Herein inner second-order it. solve an is (SOCP) proposed cone approximate program
Herein an inner second-order cone program (SOCP) approximate algorithm is proposed to solve it.
35,157
relaxation does result, computationally As a technique. algorithm heavy SDP use not our
As a result, our algorithm does not use computationally heavy SDP relaxation technique.
35,158
significance fundus doctors The retinal vessels of segmentation to diseases. for diagnose of the is
The segmentation of retinal vessels is of significance for doctors to diagnose the fundus diseases.
35,159
the mainly generator. We structure the network improve of
We mainly improve the network structure of the generator.
35,160
multiple in has enjoyed years. Reinforcement learning recent successes
Reinforcement learning has enjoyed multiple successes in recent years.
35,161
in is topic Parsimony signal active representation a research. of
Parsimony in signal representation is a topic of active research.
35,162
a conditional GAN model HiGAN \emph{high-level} of \emph{low-level} consists conditional The a GAN. and
The HiGAN model consists of a \emph{low-level} conditional GAN and a \emph{high-level} conditional GAN.
35,163
(i.e. datasets experiments on recognition Comprehensive video two challenging
Comprehensive experiments on two challenging video recognition datasets (i.e.
35,164
better? by Inspired we advancement, asked do ourselves, such can we
Inspired by such advancement, we asked ourselves, can we do better?
35,165
Networks (DCNet) propose We (DCNet++). and Networks Diverse Capsule Capsule Dense
We propose Dense Capsule Networks (DCNet) and Diverse Capsule Networks (DCNet++).
35,166
in different maps This primary forming in the incorporating layers helps capsules. learned by feature
This helps in incorporating feature maps learned by different layers in forming the primary capsules.
35,167
network, which deeper adds essentially leads to of feature discriminative DCNet, maps. convolution a learning
DCNet, essentially adds a deeper convolution network, which leads to learning of discriminative feature maps.
35,168
classification on the architectures. of efficacy using proposed Experiments image task datasets demonstrate the benchmark
Experiments on image classification task using benchmark datasets demonstrate the efficacy of the proposed architectures.
35,169
wind the forecasts. of This around on energy confidence construction aims bands work
This work aims on the construction of confidence bands around wind energy forecasts.
35,170
'open (e.g. and differing distinguish ones door', 'open cupboard'),
'open door', 'open cupboard'), and distinguish differing ones (e.g.
35,171
vs verb-only 'open labels. 'open door' using bottle')
'open door' vs 'open bottle') using verb-only labels.
35,172
legitimate ambiguities class (Fig. Current between approaches recognition and for verbs neglect action semantic overlaps
Current approaches for action recognition neglect legitimate semantic ambiguities and class overlaps between verbs (Fig.
35,173
two-stream learn multi-output fusion using regression, CNN. these as We a representations
We learn these representations as multi-output regression, using a two-stream fusion CNN.
35,174
Finally, real evaluate collected framework using data. we developed effectiveness our the of vehicle the
Finally, we evaluate the effectiveness of our developed framework using the collected real vehicle data.
35,175
wavelet identification neural nonlinear system. network online of performs the (WNN) The the
The wavelet neural network (WNN) performs the online identification of the nonlinear system.
35,176
is prove to used the theory the of The MPC. stability Lyapunov
The Lyapunov theory is used to prove the stability of the MPC.
35,177
vehicle. unmanned We control methodology identification online of an the apply the and to autonomous
We apply the methodology to the online identification and control of an unmanned autonomous vehicle.
35,178
time videos is for this longer However, consuming. very
However, for longer videos this is very time consuming.
35,179
produce accurate architectures deep datasets. Modern on highly learning segmentation challenging many semantic results
Modern deep learning architectures produce highly accurate results on many challenging semantic segmentation datasets.
35,180
neonates Limited-channel of analytic by lack research in is open, tools. Goal: accessible hindered EEG
Goal: Limited-channel EEG research in neonates is hindered by lack of open, accessible analytic tools.
35,181
three experienced by validated was The reviewers. aEEG algorithm
The aEEG algorithm was validated by three experienced reviewers.
35,182
reviewers background classification. Using methodology, a pattern assigned standard
Using standard methodology, reviewers assigned a background pattern classification.
35,183
using Pearson's coefficient. Results compared were correlation
Results were compared using Pearson's correlation coefficient.
35,184
the work the reduces manual facilitates and identity processing. of It recognition automatic
It reduces the manual work of identity recognition and facilitates the automatic processing.
35,185
the from the data. layers features learns intermediate CNN The automatically hierarchical at
The CNN learns the hierarchical features at intermediate layers automatically from the data.
35,186
does paper for performance anti-spoofing. This CNNs face of evaluation a
This paper does a performance evaluation of CNNs for face anti-spoofing.
35,187
and The CNN are used this ResNet Inception study. architectures in
The Inception and ResNet CNN architectures are used in this study.
35,188
Spoofing results are Database. Face MSU Mobile computed over The benchmark
The results are computed over benchmark MSU Mobile Face Spoofing Database.
35,189
The in for face results using these settings. different obtained favorable anti-spoofing CNN architectures are
The favorable results are obtained using these CNN architectures for face anti-spoofing in different settings.
35,190
a machine model. replacement to We this apply estimator
We apply this estimator to a machine replacement model.
35,191
to challenging Hierarchical is approach these learning a able that is principled reinforcement tackle tasks.
Hierarchical reinforcement learning is a principled approach that is able to tackle these challenging tasks.
35,192
hierarchical learning POMDP. We propose reinforcement approach a in hierarchical for learning deep
We propose a hierarchical deep reinforcement learning approach for learning in hierarchical POMDP.
35,193
The RL is hierarchical to apply MDP learning. proposed to algorithm deep POMDP and both
The deep hierarchical RL algorithm is proposed to apply to both MDP and POMDP learning.
35,194
evaluate We hierarchical challenging the algorithm proposed various POMDP. on
We evaluate the proposed algorithm on various challenging hierarchical POMDP.
35,195
tasks scene visual are parsing understanding. estimation two Depth in particularly and scene important
Depth estimation and scene parsing are two particularly important tasks in visual scene understanding.
35,196
challenging are datasets conducted two (i.e. on experiments Extensive
Extensive experiments are conducted on two challenging datasets (i.e.
35,197
novel paper method presents for correspondence. cross-domain sparse a This
This paper presents a novel method for sparse cross-domain correspondence.
35,198
input unorganized data. perspective five network of point representations the of cloud The consists
The network input consists of five perspective representations of the unorganized point cloud data.
35,199