text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
of analysis of are for massive amounts their Therefore, applications the impractical. data
Therefore, their applications for the analysis of massive amounts of data are impractical.
15,800
computations example, due For to data computationally are prohibitive. repeated accumulated
For example, repeated computations due to accumulated data are computationally prohibitive.
15,801
extension newly-arrived points. out-of-sample an We algorithm to an data which describe performs
We describe an algorithm which performs an out-of-sample extension to newly-arrived data points.
15,802
We the the prove is bounded. proposed of error algorithm that
We prove that the error of the proposed algorithm is bounded.
15,803
The similar have points a four stage. evaluated successional
The four evaluated points have a similar successional stage.
15,804
major on on shown two fronts. has Recent developments subtraction work background
Recent work on background subtraction has shown developments on two major fronts.
15,805
model from frequencies is also studied. identifiability of The parameters $k$-mer
The identifiability of model parameters from $k$-mer frequencies is also studied.
15,806
Medicago In this we model the explore issue truncatula. legume this using wild paper
In this paper we explore this issue using the wild model legume Medicago truncatula.
15,807
biomedicine breeding insight and can accelerate This research programs.
This insight can accelerate breeding and biomedicine research programs.
15,808
segmentation (FCNs). by in state-of-the-art networks currently The represented semantic convolutional is fully
The state-of-the-art in semantic segmentation is currently represented by fully convolutional networks (FCNs).
15,809
FCNs result object boundaries. around a As are produce segmentations to tend localized that poorly
As a result FCNs tend to produce segmentations that are poorly localized around object boundaries.
15,810
and segment coherence used is to The information boundary to localization. object semantic improve enhance
The boundary information is used to enhance semantic segment coherence and to improve object localization.
15,811
pairwise the boundaries energy. then We to define employ in our predicted potentials
We then employ the predicted boundaries to define pairwise potentials in our energy.
15,812
performance image tasks. its ConvNets have classification Deep shown in good
Deep ConvNets have shown its good performance in image classification tasks.
15,813
deep a remains However as recognition. it representation still action for video problem in
However it still remains as a problem in deep video representation for action recognition.
15,814
pooled the video to encoded then The VLAD with representations. descriptors form local are
The pooled local descriptors are then encoded with VLAD to form the video representations.
15,815
performance object recall. on paper of terms in focus In improving we this detection
In this paper we focus on improving object detection performance in terms of recall.
15,816
by the is sampling exploration performed This image. in proposals object
This exploration is performed by sampling object proposals in the image.
15,817
we of propose In relations objects. method a between discover addition, to higher-order groups novel
In addition, we propose a novel method to discover higher-order relations between groups of objects.
15,818
for In a scheme content we paper, intra propose this prediction screen video. novel
In this paper, we propose a novel intra prediction scheme for screen content video.
15,819
We SR techniques (i.e. benchmarks validate our on seven standard
We validate our seven techniques on standard SR benchmarks (i.e.
15,820
detection. a hand introduce new We for pipeline and fingertip localization
We introduce a new pipeline for hand localization and fingertip detection.
15,821
cities. an ozone task is of modern prediction level agencies important air quality The of
The ozone level prediction is an important task of air quality agencies of modern cities.
15,822
computer inputs based with propose output. We a system three single and
We propose a computer based system with three inputs and single output.
15,823
with This computationally machines. focus the model capable Turing modularized view the the binds with
This modularized view with the focus binds the model with the computationally capable Turing machines.
15,824
other related RAC mA+ relate contribute MADs. that we formalisms Finally, in to to
Finally, we relate mA+ to other related formalisms that contribute to RAC in MADs.
15,825
detect Moreover, to identification might be hard constraints statistically.
Moreover, identification constraints might be hard to detect statistically.
15,826
prove on the bounds of We OA in an patterns model. properties theoretical
We prove theoretical bounds on the properties of patterns in an OA model.
15,827
degree differences explained. the were Then, proposed moving existing and rate between matching
Then, the differences between proposed moving rate and existing matching degree were explained.
15,828
proposed T-S model. method identification based rate the for moving is Next, on
Next, the identification method based on moving rate is proposed for T-S model.
15,829
that significantly Test fuzzy identification. of proposed improves results show method effectiveness the the
Test results show that the proposed method significantly improves the effectiveness of fuzzy identification.
15,830
we neutrality theoretical derive the on Within NERM risks. bound a empirical generalization and framework,
Within the NERM framework, we derive a theoretical bound on empirical and generalization neutrality risks.
15,831
problem. method This attractiveness the address a prediction to paper proposes leaning challenging deep facial
This paper proposes a deep leaning method to address the challenging facial attractiveness prediction problem.
15,832
unresolved been developed limitations proposed, remain. issues have and several certain and Although methods
Although several methods have been developed and proposed, certain limitations and unresolved issues remain.
15,833
Different or measures have vectors. applied similarity these been dissimilarity to
Different similarity or dissimilarity measures have been applied to these vectors.
15,834
are their and advantages Specifically, disadvantages compared. mentioned and
Specifically, their advantages and disadvantages are mentioned and compared.
15,835
adding the model capacity competition WTA neurons. information hidden of The increases the without
The WTA competition increases the information capacity of the model without adding hidden neurons.
15,836
Logo intellectual and recognition applications, for brand protection. particularly from has property detection many images
Logo detection from images has many applications, particularly for brand recognition and intellectual property protection.
15,837
and describe a database. constructing ideas for large-scale the such We challenges
We describe the ideas and challenges for constructing such a large-scale database.
15,838
neuroscientific research humans objects. demonstrates fixed size a Recent salient have for prior that
Recent neuroscientific research demonstrates that humans have a fixed size prior for salient objects.
15,839
available source All pre-trained will at be and models GitHub. code
All source code and pre-trained models will be available at GitHub.
15,840
The calculation only advantage the has that simulations is to over executed once. be
The advantage over simulations is that the calculation has to be executed only once.
15,841
the matrix is Building automated algorithm. transition our by
Building the transition matrix is automated by our algorithm.
15,842
on a paper perform analysis we In this image automated large-scale dataset. historical
In this paper we perform automated analysis on a large-scale historical image dataset.
15,843
streaming a framework to selection. propose covariance perform We
We propose a framework to perform streaming covariance selection.
15,844
the learnt in regularization an be online efficiently This parameter allows manner. for to
This allows for the regularization parameter to be efficiently learnt in an online manner.
15,845
assumptions, non-stochastic Under in mild are results to able convergence a setting. obtain we
Under mild assumptions, we are able to obtain convergence results in a non-stochastic setting.
15,846
on work based Random a model In Field. Conditional Gaussian this new we present
In this work we present a new model based on Gaussian Conditional Random Field.
15,847
CK+ test on work two and We datasets, facial RU-FACS. our train and expression
We train and test our work on two facial expression datasets, CK+ and RU-FACS.
15,848
outperform recognition. that the Experimental approach of state art proposed evaluation our shows expression
Experimental evaluation shows that our proposed approach outperform state of the art expression recognition.
15,849
neural of systems. convolutional image networks (CNNs) the segmentation backbone semantic are state-of-art Deep
Deep convolutional neural networks (CNNs) are the backbone of state-of-art semantic image segmentation systems.
15,850
the target This an semantic optimizing system task-specific segmentation quality. end-to-end produces trainable edges in
This produces task-specific edges in an end-to-end trainable system optimizing the target semantic segmentation quality.
15,851
variational the the linearly memory unlike resources in methods. dataset with required scale size, Nevertheless,
Nevertheless, the memory resources required scale linearly with the dataset size, unlike in variational methods.
15,852
instances number This a is the large. of is very when severe limitation
This is a severe limitation when the number of instances is very large.
15,853
areas an involves problem missing of image. the The inpainting reconstructing of
The problem of inpainting involves reconstructing the missing areas of an image.
15,854
removing old damaged as many Inpainting images. such applications, photographs or has reconstructing obfuscations from
Inpainting has many applications, such as reconstructing old damaged photographs or removing obfuscations from images.
15,855
present this inpainting. for we directional algorithm paper In the diffusion
In this paper we present the directional diffusion algorithm for inpainting.
15,856
diffusion by accurately. diffusion directional The algorithm edges regular reconstructing more algorithm on the improves
The directional diffusion algorithm improves on the regular diffusion algorithm by reconstructing edges more accurately.
15,857
techniques. the Understanding ConvNets process internal commonly visualization of done is using
Understanding the internal process of ConvNets is commonly done using visualization techniques.
15,858
vital recognition, annotations semantic or for scene models Semantic object are training for segmentation understanding.
Semantic annotations are vital for training models for object recognition, semantic segmentation or scene understanding.
15,859
developmental children's in of and drawings God(s) cross-cultural other agents. and supernatural We patterns detect
We detect developmental and cross-cultural patterns in children's drawings of God(s) and other supernatural agents.
15,860
of for and experts These investigation. support the hypotheses findings questions further new raise the
These findings support the hypotheses of the experts and raise new questions for further investigation.
15,861
depth We termed instance novel a object templateNet present recognition. deep architecture based for
We present a novel deep architecture termed templateNet for depth based object instance recognition.
15,862
benefits parametrization template from layer. any this The the from additional network without
The network benefits from this without any additional parametrization from the template layer.
15,863
updates the this in to weight efficiently an needed end-to-end derive network train manner. We
We derive the weight updates needed to efficiently train this network in an end-to-end manner.
15,864
using benchmark for based We object datasets. available the two publicly depth recognition templateNet instance
We benchmark the templateNet for depth based object instance recognition using two publicly available datasets.
15,865
looking The and challenges pose distractors. large multiple clutter, present similar datasets of variations
The datasets present multiple challenges of clutter, large pose variations and similar looking distractors.
15,866
of based alpha method minimization approximate the on is inference Black-box (BB-$\alpha$) new a $\alpha$-divergences.
Black-box alpha (BB-$\alpha$) is a new approximate inference method based on the minimization of $\alpha$-divergences.
15,867
it gradient because datasets large to scales implemented be can using BB-$\alpha$ stochastic descent.
BB-$\alpha$ scales to large datasets because it can be implemented using stochastic gradient descent.
15,868
easily gradients differentiation. obtained can be automatic These using
These gradients can be easily obtained using automatic differentiation.
15,869
approaches, recent video the three Compared representation makes contributions. this inference paper following to
Compared to recent video representation inference approaches, this paper makes the following three contributions.
15,870
operations attaining while Second, lessened non-linearity. are significantly computation more
Second, computation operations are significantly lessened while attaining more non-linearity.
15,871
apply information new to method where temporal video the role. a We plays captioning crucial
We apply the new method to video captioning where temporal information plays a crucial role.
15,872
benchmarks. outperforms our that method on the captioning video Experiments demonstrate state-of-the-art
Experiments demonstrate that our method outperforms the state-of-the-art on video captioning benchmarks.
15,873
neural predictions functions. most provides fastest while the networks accurate metamodel yield the Kriging
Kriging provides the most accurate predictions while neural networks yield the fastest metamodel functions.
15,874
computation All failures. strong metamodels three can detect conveniently
All three metamodels can conveniently detect strong computation failures.
15,875
most find we instability code Kriging tools. provides detection, that useful For the
For code instability detection, we find that Kriging provides the most useful tools.
15,876
time consuming. turns be out quite process the to Thus,
Thus, the process turns out to be quite time consuming.
15,877
above to solution simple approach discussed. novel problem a and Here the is a
Here a novel approach and a simple solution to the above problem is discussed.
15,878
novel based classification propose this In convolutional architecture on ProNet we networks. neural paper, a
In this paper, we propose a novel classification architecture ProNet based on convolutional neural networks.
15,879
We minimize formulated the to method proximal-subgradient use accelerated function. an cost
We use an accelerated proximal-subgradient method to minimize the formulated cost function.
15,880
of its prove analyze and the convergence. the performance We proposed algorithm
We analyze the performance of the proposed algorithm and prove its convergence.
15,881
methods the A novel heterogeneous is fusion approach detection object for proposed. of
A novel approach for the fusion of heterogeneous object detection methods is proposed.
15,882
create spatially to Forest extends (SCRF) Random Spatially Forest Random labeling. Coherent coherent
Spatially Coherent Random Forest (SCRF) extends Random Forest to create spatially coherent labeling.
15,883
in We can SCRF settings that be used believe as other well.
We believe that SCRF can be used in other settings as well.
15,884
which many of in are that remains from there However, far still humans. performance cases
However, there are still many cases in which performance remains far from that of humans.
15,885
allows our network classify even deformations. under to intraclass adoption features samples precisely This
This features adoption allows our network to classify precisely intraclass samples even under deformations.
15,886
should long-term attractive campaigns effectively. To organized be managed these more deposit and overcome problems,
To overcome these problems, attractive long-term deposit campaigns should be organized and managed more effectively.
15,887
applications. in search for vision hashing methods widely-used neighbor Supervised computer nearest are
Supervised hashing methods are widely-used for nearest neighbor search in computer vision applications.
15,888
state-of-the-art batch-learners. hashing Most employ supervised approaches
Most state-of-the-art supervised hashing approaches employ batch-learners.
15,889
large batch-learning datasets. when be confronted inefficient can Unfortunately, training strategies with
Unfortunately, batch-learning strategies can be inefficient when confronted with large training datasets.
15,890
the with online algorithm, approach our an offers size. complexity is dataset linear it Since
Since it is an online algorithm, our approach offers linear complexity with the dataset size.
15,891
to we hash table reduce Thus, a framework also updates. propose
Thus, we also propose a framework to reduce hash table updates.
15,892
networks. of of focus The learning Bayesian these paper this scalable is approximate
The focus of this paper is scalable approximate Bayesian learning of these networks.
15,893
pathways transduction disorders, variety cancer. lead signal can including complex to a Dysregulation of in
Dysregulation in signal transduction pathways can lead to a variety of complex disorders, including cancer.
15,894
Each mutations. of components gain-of-function both loss-of-function under perturbed and model's the was
Each of the model's components was perturbed under both loss-of-function and gain-of-function mutations.
15,895
the of paying both However, and non-dispersal. entails costs dispersal this
However, this entails paying the costs of both dispersal and non-dispersal.
15,896
approximate their to of distributions are methods broadly or posterior Importance moments. used Sampling some
Importance Sampling methods are broadly used to approximate posterior distributions or some of their moments.
15,897
single approach, In proposal weighted a distribution standard samples properly. from are drawn and its
In its standard approach, samples are drawn from a single proposal distribution and weighted properly.
15,898
a formulate ZSR We problem. prediction binary as
We formulate ZSR as a binary prediction problem.
15,899