Unnamed: 0.1
int64
0
41k
Unnamed: 0
int64
0
41k
author
stringlengths
9
1.39k
id
stringlengths
11
18
summary
stringlengths
25
3.66k
title
stringlengths
4
258
year
int64
1.99k
2.02k
arxiv_url
stringlengths
32
39
info
stringlengths
523
3.18k
embeddings
stringlengths
16.9k
17.1k
600
600
['Lili Mou', 'Rui Yan', 'Ge Li', 'Lu Zhang', 'Zhi Jin']
1512.06612v2
Recent language models, especially those based on recurrent neural networks (RNNs), make it possible to generate natural language from a learned probability. Language generation has wide applications including machine translation, summarization, question answering, conversation systems, etc. Existing methods typically ...
Backward and Forward Language Modeling for Constrained Sentence Generation
2,015
http://arxiv.org/pdf/1512.06612v2
Title Backward Forward Language Modeling Constrained Sentence Generation Summary Recent language model especially based recurrent neural network RNNs make possible generate natural language learned probability Language generation wide application including machine translation summarization question answering conversati...
[0.0713164359331131, 0.04077165201306343, -0.00351351173594594, 0.009156038984656334, -0.050216205418109894, -0.03484518453478813, -0.01834772154688835, 0.015102071687579155, -0.011920073069632053, -0.04726139456033707, 0.04541974887251854, -0.043521977961063385, 0.005748002789914608, 0.04906434938311577, 0.02093552052...
601
601
['Kyuyeon Hwang', 'Minjae Lee', 'Wonyong Sung']
1512.08903v1
In this paper, we propose a context-aware keyword spotting model employing a character-level recurrent neural network (RNN) for spoken term detection in continuous speech. The RNN is end-to-end trained with connectionist temporal classification (CTC) to generate the probabilities of character and word-boundary labels. ...
Online Keyword Spotting with a Character-Level Recurrent Neural Network
2,015
http://arxiv.org/pdf/1512.08903v1
Title Online Keyword Spotting CharacterLevel Recurrent Neural Network Summary paper propose contextaware keyword spotting model employing characterlevel recurrent neural network RNN spoken term detection continuous speech RNN endtoend trained connectionist temporal classification CTC generate probability character word...
[-0.006053483579307795, 0.017226427793502808, 0.04451555758714676, 0.04900750890374184, 0.017928095534443855, -0.02497158572077751, -0.008989205583930016, 0.008729066699743271, -0.047539252787828445, -0.03288761526346207, -0.03752556070685387, -0.04293350502848625, 0.030366219580173492, 0.09057103842496872, 0.001829272...
602
602
['Zhenhao Ge', 'Yufang Sun']
1602.07393v1
Authorship attribution refers to the task of automatically determining the author based on a given sample of text. It is a problem with a long history and has a wide range of application. Building author profiles using language models is one of the most successful methods to automate this task. New language modeling me...
Domain Specific Author Attribution Based on Feedforward Neural Network Language Models
2,016
http://arxiv.org/pdf/1602.07393v1
Title Domain Specific Author Attribution Based Feedforward Neural Network Language Models Summary Authorship attribution refers task automatically determining author based given sample text problem long history wide range application Building author profile using language model one successful method automate task New l...
[0.049107592552900314, 0.05301670730113983, -0.03609152138233185, 0.024805482476949692, -0.05954847112298012, -0.019497159868478775, 0.10058499872684479, -0.013518590480089188, 0.01585269533097744, -0.027639785781502724, 0.015150275081396103, -0.0017323875799775124, 0.024015109986066818, 0.03368305042386055, 0.02501071...
603
603
['Liang Lu', 'Lingpeng Kong', 'Chris Dyer', 'Noah A. Smith', 'Steve Renals']
1603.00223v2
We study the segmental recurrent neural network for end-to-end acoustic modelling. This model connects the segmental conditional random field (CRF) with a recurrent neural network (RNN) used for feature extraction. Compared to most previous CRF-based acoustic models, it does not rely on an external system to provide fe...
Segmental Recurrent Neural Networks for End-to-end Speech Recognition
2,016
http://arxiv.org/pdf/1603.00223v2
Title Segmental Recurrent Neural Networks Endtoend Speech Recognition Summary study segmental recurrent neural network endtoend acoustic modelling model connects segmental conditional random field CRF recurrent neural network RNN used feature extraction Compared previous CRFbased acoustic model rely external system pro...
[0.01909462735056877, 0.029673578217625618, 0.037180881947278976, 0.03395266830921173, 0.002734440378844738, -0.03491126373410225, -0.0002440836833557114, 0.007297059986740351, -0.08068285882472992, -0.017752185463905334, -0.02081473171710968, -0.013872994109988213, 0.03581831231713295, 0.017155731096863747, -0.0233709...
604
604
['Lili Mou', 'Zhao Meng', 'Rui Yan', 'Ge Li', 'Yan Xu', 'Lu Zhang', 'Zhi Jin']
1603.06111v2
Transfer learning is aimed to make use of valuable knowledge in a source domain to help model performance in a target domain. It is particularly important to neural networks, which are very likely to be overfitting. In some fields like image processing, many studies have shown the effectiveness of neural network-based ...
How Transferable are Neural Networks in NLP Applications?
2,016
http://arxiv.org/pdf/1603.06111v2
Title Transferable Neural Networks NLP Applications Summary Transfer learning aimed make use valuable knowledge source domain help model performance target domain particularly important neural network likely overfitting field like image processing many study shown effectiveness neural networkbased transfer learning neu...
[0.0072393398731946945, 0.044481221586465836, -0.004846601746976376, 0.06205492094159126, -0.013320247642695904, -0.012743615545332432, 0.0037230942398309708, 0.002024505054578185, -0.02883164770901203, -0.030240140855312347, -0.04762844368815422, 0.05009543523192406, -0.008013979531824589, 0.06258271634578705, 0.02063...
605
605
['Wei-Ning Hsu', 'Yu Zhang', 'James Glass']
1603.07044v1
We apply a general recurrent neural network (RNN) encoder framework to community question answering (cQA) tasks. Our approach does not rely on any linguistic processing, and can be applied to different languages or domains. Further improvements are observed when we extend the RNN encoders with a neural attention mechan...
Recurrent Neural Network Encoder with Attention for Community Question Answering
2,016
http://arxiv.org/pdf/1603.07044v1
Title Recurrent Neural Network Encoder Attention Community Question Answering Summary apply general recurrent neural network RNN encoder framework community question answering cQA task approach rely linguistic processing applied different language domain improvement observed extend RNN encoders neural attention mechani...
[0.028209848329424858, 0.039999667555093765, -0.011160486377775669, 0.06346011161804199, -0.007395118474960327, -0.006305999588221312, 0.0022823307663202286, -0.00979146920144558, -0.008441049605607986, -0.039311543107032776, 0.01076226495206356, 0.0029576115775853395, 0.0028488426469266415, 0.030060268938541412, -0.00...
606
606
['Saurabh Kataria']
1603.07646v1
We consider the problem of learning distributed representations for tags from their associated content for the task of tag recommendation. Considering tagging information is usually very sparse, effective learning from content and tag association is very crucial and challenging task. Recently, various neural representa...
Recursive Neural Language Architecture for Tag Prediction
2,016
http://arxiv.org/pdf/1603.07646v1
Title Recursive Neural Language Architecture Tag Prediction Summary consider problem learning distributed representation tag associated content task tag recommendation Considering tagging information usually sparse effective learning content tag association crucial challenging task Recently various neural representatio...
[0.046709440648555756, -0.005000155884772539, -0.046266429126262665, 0.02858760952949524, 0.001003048848360777, 0.0013018466997891665, -0.013201013207435608, -0.004298039246350527, 0.05208227410912514, -0.023654785007238388, -0.025690404698252678, -0.0348496213555336, 0.024195024743676186, 0.07179399579763412, 0.014678...
607
607
['Rohit Prabhavalkar', 'Ouais Alsharif', 'Antoine Bruguier', 'Ian McGraw']
1603.08042v2
We study the problem of compressing recurrent neural networks (RNNs). In particular, we focus on the compression of RNN acoustic models, which are motivated by the goal of building compact and accurate speech recognition systems which can be run efficiently on mobile devices. In this work, we present a technique for ge...
On the Compression of Recurrent Neural Networks with an Application to LVCSR acoustic modeling for Embedded Speech Recognition
2,016
http://arxiv.org/pdf/1603.08042v2
Title Compression Recurrent Neural Networks Application LVCSR acoustic modeling Embedded Speech Recognition Summary study problem compressing recurrent neural network RNNs particular focus compression RNN acoustic model motivated goal building compact accurate speech recognition system run efficiently mobile device wor...
[-0.01828327216207981, -0.004037776961922646, 0.015006584115326405, 0.054580144584178925, -0.012824430130422115, -0.023773901164531708, 0.027628332376480103, 0.022657198831439018, -0.06416162848472595, 0.023214837536215782, -0.03965678811073303, -0.013697794638574123, 0.04727723076939583, -0.015872197225689888, 0.00747...
608
608
['Caglar Gulcehre', 'Sungjin Ahn', 'Ramesh Nallapati', 'Bowen Zhou', 'Yoshua Bengio']
1603.08148v3
The problem of rare and unknown words is an important issue that can potentially influence the performance of many NLP systems, including both the traditional count-based and the deep learning models. We propose a novel way to deal with the rare and unseen words for the neural network models using attention. Our model ...
Pointing the Unknown Words
2,016
http://arxiv.org/pdf/1603.08148v3
Title Pointing Unknown Words Summary problem rare unknown word important issue potentially influence performance many NLP system including traditional countbased deep learning model propose novel way deal rare unseen word neural network model using attention model us two softmax layer order predict next word conditiona...
[0.0698966532945633, 0.012464814819395542, 0.01565462164580822, 0.032125454396009445, 0.0012413514778017998, 0.0022937203757464886, -0.005260268691927195, 0.0024114649277180433, -0.021239766851067543, -0.04228523001074791, 0.010414793156087399, -0.006894045043736696, 0.017795942723751068, 0.015037612058222294, 0.041376...
609
609
['Zhenyao Zhu', 'Jesse H. Engel', 'Awni Hannun']
1603.09509v2
Deep learning has dramatically improved the performance of speech recognition systems through learning hierarchies of features optimized for the task at hand. However, true end-to-end learning, where features are learned directly from waveforms, has only recently reached the performance of hand-tailored representations...
Learning Multiscale Features Directly From Waveforms
2,016
http://arxiv.org/pdf/1603.09509v2
Title Learning Multiscale Features Directly Waveforms Summary Deep learning dramatically improved performance speech recognition system learning hierarchy feature optimized task hand However true endtoend learning feature learned directly waveform recently reached performance handtailored representation based Fourier t...
[-0.033016979694366455, 0.059531424194574356, 0.01716112531721592, 0.058574970811605453, 0.017842482775449753, 5.215239434619434e-05, 0.06738042831420898, 0.010663111694157124, -0.023073263466358185, 0.011679023504257202, -0.08431396633386612, -0.05607355386018753, 0.037901535630226135, 0.06979480385780334, 0.009736104...
610
610
['Petr Baudis', 'Silvestr Stanko', 'Jan Sedivy']
1605.04655v2
We consider the problem of Recognizing Textual Entailment within an Information Retrieval context, where we must simultaneously determine the relevancy as well as degree of entailment for individual pieces of evidence to determine a yes/no answer to a binary natural language question. We compare several variants of n...
Joint Learning of Sentence Embeddings for Relevance and Entailment
2,016
http://arxiv.org/pdf/1605.04655v2
Title Joint Learning Sentence Embeddings Relevance Entailment Summary consider problem Recognizing Textual Entailment within Information Retrieval context must simultaneously determine relevancy well degree entailment individual piece evidence determine yesno answer binary natural language question compare several vari...
[0.06912939995527267, 0.0009819197002798319, -0.000473639986012131, 0.045569971203804016, -0.05108325555920601, 0.03289928287267685, -0.02245612069964409, -0.008449180983006954, 0.021231716498732567, -0.041814446449279785, 0.0102603854611516, 0.051032230257987976, -0.028879690915346146, 0.03999879211187363, -0.00059553...
611
611
['Xiaodong Gu', 'Hongyu Zhang', 'Dongmei Zhang', 'Sunghun Kim']
1605.08535v3
Developers often wonder how to implement a certain functionality (e.g., how to parse XML files) using APIs. Obtaining an API usage sequence based on an API-related natural language query is very helpful in this regard. Given a query, existing approaches utilize information retrieval models to search for matching API se...
Deep API Learning
2,016
http://arxiv.org/pdf/1605.08535v3
Title Deep API Learning Summary Developers often wonder implement certain functionality eg parse XML file using APIs Obtaining API usage sequence based APIrelated natural language query helpful regard Given query existing approach utilize information retrieval model search matching API sequence approach treat query API...
[0.022035649046301842, 0.08039530366659164, -0.022920427843928337, 0.05839980021119118, -0.023591330274939537, -0.025041567161679268, -0.018716687336564064, 0.004967098589986563, 0.017906080931425095, -0.03410175442695618, 0.015619850717484951, 0.014092438854277134, 0.019767960533499718, 0.09026051312685013, 0.00809093...
612
612
['Ozan Caglayan', 'Walid Aransa', 'Yaxing Wang', 'Marc Masana', 'Mercedes García-Martínez', 'Fethi Bougares', 'Loïc Barrault', 'Joost van de Weijer']
1605.09186v4
This paper presents the systems developed by LIUM and CVC for the WMT16 Multimodal Machine Translation challenge. We explored various comparative methods, namely phrase-based systems and attentional recurrent neural networks models trained using monomodal or multimodal data. We also performed a human evaluation in orde...
Does Multimodality Help Human and Machine for Translation and Image Captioning?
2,016
http://arxiv.org/pdf/1605.09186v4
Title Multimodality Help Human Machine Translation Image Captioning Summary paper present system developed LIUM CVC WMT16 Multimodal Machine Translation challenge explored various comparative method namely phrasebased system attentional recurrent neural network model trained using monomodal multimodal data also perform...
[0.0539010725915432, 0.0076685454696416855, 0.01904091238975525, 0.06651923060417175, -0.01815539225935936, 0.015360726043581963, 0.03222649171948433, -0.017729010432958603, -0.015600577928125858, -0.06154349073767662, -0.04799551144242287, -0.04930960759520531, 0.044641319662332535, 0.01870400458574295, 0.051146630197...
613
613
['Alexis Conneau', 'Holger Schwenk', 'Loïc Barrault', 'Yann Lecun']
1606.01781v2
The dominant approach for many NLP tasks are recurrent neural networks, in particular LSTMs, and convolutional neural networks. However, these architectures are rather shallow in comparison to the deep convolutional networks which have pushed the state-of-the-art in computer vision. We present a new architecture (VDCNN...
Very Deep Convolutional Networks for Text Classification
2,016
http://arxiv.org/pdf/1606.01781v2
Title Deep Convolutional Networks Text Classification Summary dominant approach many NLP task recurrent neural network particular LSTMs convolutional neural network However architecture rather shallow comparison deep convolutional network pushed stateoftheart computer vision present new architecture VDCNN text processi...
[0.060872286558151245, 0.024899516254663467, 0.0042010704055428505, 0.07925429940223694, -0.017536798492074013, 0.012691543437540531, 0.03025246411561966, 0.02713078074157238, 0.017733626067638397, -0.07143654674291611, -0.014239005744457245, -0.006598727311939001, -0.004835042171180248, 0.044520456343889236, -0.008369...
614
614
['Marco Dinarelli', 'Isabelle Tellier']
1606.02555v1
In this paper we study different types of Recurrent Neural Networks (RNN) for sequence labeling tasks. We propose two new variants of RNNs integrating improvements for sequence labeling, and we compare them to the more traditional Elman and Jordan RNNs. We compare all models, either traditional or new, on four distinct...
Improving Recurrent Neural Networks For Sequence Labelling
2,016
http://arxiv.org/pdf/1606.02555v1
Title Improving Recurrent Neural Networks Sequence Labelling Summary paper study different type Recurrent Neural Networks RNN sequence labeling task propose two new variant RNNs integrating improvement sequence labeling compare traditional Elman Jordan RNNs compare model either traditional new four distinct task sequen...
[0.04753683879971504, -0.0005172011442482471, 0.02614475227892399, 0.044264134019613266, -0.04098315164446831, -0.0019576717168092728, 0.0019845752976834774, -0.02557271532714367, -0.016920369118452072, -0.029783299192786217, -0.01887935772538185, -0.037809696048498154, 0.026239173486828804, 0.026845987886190414, 0.013...
615
615
['Marek Rei', 'Ronan Cummins']
1606.03144v1
We investigate the task of assessing sentence-level prompt relevance in learner essays. Various systems using word overlap, neural embeddings and neural compositional models are evaluated on two datasets of learner writing. We propose a new method for sentence-level similarity calculation, which learns to adjust the we...
Sentence Similarity Measures for Fine-Grained Estimation of Topical Relevance in Learner Essays
2,016
http://arxiv.org/pdf/1606.03144v1
Title Sentence Similarity Measures FineGrained Estimation Topical Relevance Learner Essays Summary investigate task assessing sentencelevel prompt relevance learner essay Various system using word overlap neural embeddings neural compositional model evaluated two datasets learner writing propose new method sentenceleve...
[0.06661763787269592, 0.027093369513750076, 0.0011273910058662295, 0.0300119798630476, -0.040839653462171555, 0.011714022606611252, 0.019872717559337616, -0.019970372319221497, -0.0007985280826687813, -0.04079223424196243, -0.019642191007733345, -0.009494829922914505, 0.015323278494179249, 0.013418403454124928, -0.0068...
616
616
['Hwaran Lee', 'Geonmin Kim', 'Ho-Gyeong Kim', 'Sang-Hoon Oh', 'Soo-Young Lee']
1606.03207v2
Convolutional neural networks (CNNs) with convolutional and pooling operations along the frequency axis have been proposed to attain invariance to frequency shifts of features. However, this is inappropriate with regard to the fact that acoustic features vary in frequency. In this paper, we contend that convolution alo...
Deep CNNs along the Time Axis with Intermap Pooling for Robustness to Spectral Variations
2,016
http://arxiv.org/pdf/1606.03207v2
Title Deep CNNs along Time Axis Intermap Pooling Robustness Spectral Variations Summary Convolutional neural network CNNs convolutional pooling operation along frequency axis proposed attain invariance frequency shift feature However inappropriate regard fact acoustic feature vary frequency paper contend convolution al...
[-0.006704570259898901, -0.0024397079832851887, 0.002484459662809968, 0.047621820122003555, -0.01375389751046896, -0.014747148379683495, 0.04437816143035889, 0.00294478889554739, -0.07536257058382034, 0.023285526782274246, -0.07044918090105057, -0.013373246416449547, 0.03883817791938782, 0.020465411245822906, 0.0048978...
617
617
['Dimitrios Alikaniotis', 'Helen Yannakoudakis', 'Marek Rei']
1606.04289v2
Automated Text Scoring (ATS) provides a cost-effective and consistent alternative to human marking. However, in order to achieve good performance, the predictive features of the system need to be manually engineered by human experts. We introduce a model that forms word representations by learning the extent to which s...
Automatic Text Scoring Using Neural Networks
2,016
http://arxiv.org/pdf/1606.04289v2
Title Automatic Text Scoring Using Neural Networks Summary Automated Text Scoring ATS provides costeffective consistent alternative human marking However order achieve good performance predictive feature system need manually engineered human expert introduce model form word representation learning extent specific word ...
[0.03240957856178284, -0.006253769621253014, -0.019459666684269905, 0.02273603342473507, -0.02264132909476757, -0.011064781807363033, 0.018832484260201454, 0.024143334478139877, 0.0169227197766304, -0.059367869049310684, -0.016813043504953384, 0.010927054099738598, 0.061029329895973206, 0.0573597326874733, 0.0074659073...
618
618
['Albert Zeyer', 'Patrick Doetsch', 'Paul Voigtlaender', 'Ralf Schlüter', 'Hermann Ney']
1606.06871v2
We present a comprehensive study of deep bidirectional long short-term memory (LSTM) recurrent neural network (RNN) based acoustic models for automatic speech recognition (ASR). We study the effect of size and depth and train models of up to 8 layers. We investigate the training aspect and study different variants of o...
A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition
2,016
http://arxiv.org/pdf/1606.06871v2
Title Comprehensive Study Deep Bidirectional LSTM RNNs Acoustic Modeling Speech Recognition Summary present comprehensive study deep bidirectional long shortterm memory LSTM recurrent neural network RNN based acoustic model automatic speech recognition ASR study effect size depth train model 8 layer investigate trainin...
[0.018957188352942467, -0.0008593597449362278, 0.02219080552458763, 0.09828613698482513, -0.019801892340183258, -0.0029993506614118814, 0.02846546471118927, -0.02588002383708954, -0.02469421923160553, -0.025839192792773247, -0.06362873315811157, -0.053315840661525726, 0.0055857026018202305, 0.011284314095973969, 0.0029...
619
619
['Yoon Kim', 'Alexander M. Rush']
1606.07947v4
Neural machine translation (NMT) offers a novel alternative formulation of translation that is potentially simpler than statistical approaches. However to reach competitive performance, NMT models need to be exceedingly large. In this paper we consider applying knowledge distillation approaches (Bucila et al., 2006; Hi...
Sequence-Level Knowledge Distillation
2,016
http://arxiv.org/pdf/1606.07947v4
Title SequenceLevel Knowledge Distillation Summary Neural machine translation NMT offer novel alternative formulation translation potentially simpler statistical approach However reach competitive performance NMT model need exceedingly large paper consider applying knowledge distillation approach Bucila et al 2006 Hint...
[0.042545102536678314, 0.07262073457241058, 0.002678351476788521, 0.030915237963199615, -0.016466815024614334, -0.013895583339035511, 0.00313385808840394, 0.009221713058650494, -0.07044472545385361, -0.038467105478048325, 0.016061589121818542, -0.0013897559838369489, 0.025277916342020035, 0.02924548089504242, 0.0240577...
620
620
['Madhusudan Lakshmana', 'Sundararajan Sellamanickam', 'Shirish Shevade', 'Keerthi Selvaraj']
1608.00466v2
The state-of-the-art CNN models give good performance on sentence classification tasks. The purpose of this work is to empirically study desirable properties such as semantic coherence, attention mechanism and reusability of CNNs in these tasks. Semantically coherent kernels are preferable as they are a lot more interp...
Learning Semantically Coherent and Reusable Kernels in Convolution Neural Nets for Sentence Classification
2,016
http://arxiv.org/pdf/1608.00466v2
Title Learning Semantically Coherent Reusable Kernels Convolution Neural Nets Sentence Classification Summary stateoftheart CNN model give good performance sentence classification task purpose work empirically study desirable property semantic coherence attention mechanism reusability CNNs task Semantically coherent ke...
[0.028136268258094788, 0.0006274341722019017, 0.009030177257955074, 0.08677586168050766, 9.5728573796805e-05, -0.03109797276556492, -0.007404872216284275, -0.012728176079690456, -0.003021592739969492, -0.06996281445026398, 0.011573974043130875, 0.06878609210252762, -0.01573127880692482, 0.03351372852921486, 0.001849544...
621
621
['Patrick Doetsch', 'Albert Zeyer', 'Paul Voigtlaender', 'Ilya Kulikov', 'Ralf Schlüter', 'Hermann Ney']
1608.00895v2
In this work we release our extensible and easily configurable neural network training software. It provides a rich set of functional layers with a particular focus on efficient training of recurrent neural network topologies on multiple GPUs. The source of the software package is public and freely available for academ...
RETURNN: The RWTH Extensible Training framework for Universal Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1608.00895v2
Title RETURNN RWTH Extensible Training framework Universal Recurrent Neural Networks Summary work release extensible easily configurable neural network training software provides rich set functional layer particular focus efficient training recurrent neural network topology multiple GPUs source software package public ...
[-0.0106066158041358, 0.029567375779151917, 0.012187649495899677, 0.03967103362083435, -0.01368166133761406, -0.000895736156962812, 0.040856894105672836, 0.004182773642241955, -0.016622059047222137, -0.004811123013496399, -0.03874160349369049, -0.042976900935173035, 0.032456908375024796, 0.055932242423295975, 0.0137951...
622
622
['Kyuyeon Hwang', 'Wonyong Sung']
1609.03777v2
Recurrent neural network (RNN) based character-level language models (CLMs) are extremely useful for modeling out-of-vocabulary words by nature. However, their performance is generally much worse than the word-level language models (WLMs), since CLMs need to consider longer history of tokens to properly predict the nex...
Character-Level Language Modeling with Hierarchical Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1609.03777v2
Title CharacterLevel Language Modeling Hierarchical Recurrent Neural Networks Summary Recurrent neural network RNN based characterlevel language model CLMs extremely useful modeling outofvocabulary word nature However performance generally much worse wordlevel language model WLMs since CLMs need consider longer history...
[0.03231010586023331, 0.066754549741745, 0.01192980632185936, 0.039426058530807495, -0.03793578967452049, 0.006929038558155298, 0.048987630754709244, 0.01880742609500885, -0.0144269410520792, -0.035498570650815964, -0.007920471951365471, -0.05754922702908516, 0.032995760440826416, 0.03209879249334335, 0.004534644540399...
623
623
['Zhiyuan Tang', 'Lantian Li', 'Dong Wang']
1609.08337v1
Research on multilingual speech recognition remains attractive yet challenging. Recent studies focus on learning shared structures under the multi-task paradigm, in particular a feature sharing structure. This approach has been found effective to improve performance on each individual language. However, this approach i...
Multi-task Recurrent Model for True Multilingual Speech Recognition
2,016
http://arxiv.org/pdf/1609.08337v1
Title Multitask Recurrent Model True Multilingual Speech Recognition Summary Research multilingual speech recognition remains attractive yet challenging Recent study focus learning shared structure multitask paradigm particular feature sharing structure approach found effective improve performance individual language H...
[0.011886225081980228, -0.02644840069115162, 0.0017623017774894834, 0.07945913821458817, -0.020982034504413605, 0.03868953511118889, 0.042908940464258194, -0.020795507356524467, 0.008639167994260788, -0.023615295067429543, -0.09359521418809891, -0.06473927199840546, 0.04473632946610451, 0.0031332438811659813, -0.006202...
624
624
['A. Hassan', 'M. R. Amin', 'N. Mohammed', 'A. K. A. Azad']
1610.00369v2
Sentiment Analysis (SA) is an action research area in the digital age. With rapid and constant growth of online social media sites and services, and the increasing amount of textual data such as - statuses, comments, reviews etc. available in them, application of automatic SA is on the rise. However, most of the resear...
Sentiment Analysis on Bangla and Romanized Bangla Text (BRBT) using Deep Recurrent models
2,016
http://arxiv.org/pdf/1610.00369v2
Title Sentiment Analysis Bangla Romanized Bangla Text BRBT using Deep Recurrent model Summary Sentiment Analysis SA action research area digital age rapid constant growth online social medium site service increasing amount textual data status comment review etc available application automatic SA rise However research w...
[0.05192713811993599, 0.027055339887738228, -0.0023078983649611473, 0.05712750926613808, -0.03185954689979553, 0.005102986469864845, -0.018143607303500175, 0.008185272105038166, -0.0019556041806936264, -0.05489837005734444, -0.005273478105664253, -0.059093259274959564, 0.01299042534083128, 0.07535676658153534, -0.02079...
625
625
['Marek Rei', 'Gamal K. O. Crichton', 'Sampo Pyysalo']
1611.04361v1
Sequence labeling architectures use word embeddings for capturing similarity, but suffer when handling previously unseen or rare words. We investigate character-level extensions to such models and propose a novel architecture for combining alternative word representations. By using an attention mechanism, the model is ...
Attending to Characters in Neural Sequence Labeling Models
2,016
http://arxiv.org/pdf/1611.04361v1
Title Attending Characters Neural Sequence Labeling Models Summary Sequence labeling architecture use word embeddings capturing similarity suffer handling previously unseen rare word investigate characterlevel extension model propose novel architecture combining alternative word representation using attention mechanism...
[0.061423782259225845, 0.026141155511140823, 0.010138444602489471, 0.027237990871071815, 0.009262291714549065, 0.022990919649600983, 0.0010185353457927704, -0.012745029293000698, 0.026322072371840477, -0.007542262319475412, -0.021471505984663963, -0.05359182134270668, 0.028383905068039894, 0.04412228241562843, 0.049286...
626
626
['Volkan Cirik', 'Eduard Hovy', 'Louis-Philippe Morency']
1611.06204v1
Curriculum Learning emphasizes the order of training instances in a computational learning setup. The core hypothesis is that simpler instances should be learned early as building blocks to learn more complex ones. Despite its usefulness, it is still unknown how exactly the internal representation of models are affecte...
Visualizing and Understanding Curriculum Learning for Long Short-Term Memory Networks
2,016
http://arxiv.org/pdf/1611.06204v1
Title Visualizing Understanding Curriculum Learning Long ShortTerm Memory Networks Summary Curriculum Learning emphasizes order training instance computational learning setup core hypothesis simpler instance learned early building block learn complex one Despite usefulness still unknown exactly internal representation ...
[0.003692563623189926, -0.04951534792780876, -0.04816649109125137, 0.01879819668829441, 0.009423155337572098, 0.03182114660739899, 0.018345246091485023, -0.05829724669456482, -0.017054418101906776, -0.014998605474829674, -0.0012486756313592196, 0.014945280738174915, 0.03155054897069931, 0.0015359147218987346, 0.0460324...
627
627
['Tom Sercu', 'Vaibhava Goel']
1611.09288v2
In computer vision pixelwise dense prediction is the task of predicting a label for each pixel in the image. Convolutional neural networks achieve good performance on this task, while being computationally efficient. In this paper we carry these ideas over to the problem of assigning a sequence of labels to a set of sp...
Dense Prediction on Sequences with Time-Dilated Convolutions for Speech Recognition
2,016
http://arxiv.org/pdf/1611.09288v2
Title Dense Prediction Sequences TimeDilated Convolutions Speech Recognition Summary computer vision pixelwise dense prediction task predicting label pixel image Convolutional neural network achieve good performance task computationally efficient paper carry idea problem assigning sequence label set speech frame task c...
[-0.01365703996270895, 0.03957163169980049, 0.006407111417502165, 0.04693054407835007, -0.008529930375516415, -0.023427637293934822, 0.06719914823770523, 0.002023087115958333, -0.053115736693143845, -0.005807191599160433, -0.014297999441623688, -0.03579714149236679, 0.032626617699861526, 0.06886547803878784, 0.01784059...
628
628
['Kartik Audhkhasi', 'Andrew Rosenberg', 'Abhinav Sethy', 'Bhuvana Ramabhadran', 'Brian Kingsbury']
1701.04313v1
End-to-end (E2E) systems have achieved competitive results compared to conventional hybrid hidden Markov model (HMM)-deep neural network based automatic speech recognition (ASR) systems. Such E2E systems are attractive due to the lack of dependence on alignments between input acoustic and output grapheme or HMM state s...
End-to-End ASR-free Keyword Search from Speech
2,017
http://arxiv.org/pdf/1701.04313v1
Title EndtoEnd ASRfree Keyword Search Speech Summary Endtoend E2E system achieved competitive result compared conventional hybrid hidden Markov model HMMdeep neural network based automatic speech recognition ASR system E2E system attractive due lack dependence alignment input acoustic output grapheme HMM state sequence...
[0.019321588799357414, 0.02940099500119686, 0.027120331302285194, 0.0550609789788723, -0.006799870170652866, -0.020971085876226425, 0.003670979058369994, 0.0041054971516132355, -0.023339271545410156, -0.05908973887562752, -0.03312704339623451, 0.004711388610303402, -0.012850308790802956, 0.06760329753160477, 0.00777878...
629
629
['Sam Wiseman', 'Sumit Chopra', "Marc'Aurelio Ranzato", 'Arthur Szlam', 'Ruoyu Sun', 'Soumith Chintala', 'Nicolas Vasilache']
1702.04770v1
While Truncated Back-Propagation through Time (BPTT) is the most popular approach to training Recurrent Neural Networks (RNNs), it suffers from being inherently sequential (making parallelization difficult) and from truncating gradient flow between distant time-steps. We investigate whether Target Propagation (TPROP) s...
Training Language Models Using Target-Propagation
2,017
http://arxiv.org/pdf/1702.04770v1
Title Training Language Models Using TargetPropagation Summary Truncated BackPropagation Time BPTT popular approach training Recurrent Neural Networks RNNs suffers inherently sequential making parallelization difficult truncating gradient flow distant timesteps investigate whether Target Propagation TPROP style approac...
[0.022913193330168724, 0.018226375803351402, -0.0020813068840652704, 0.01972082257270813, -0.017786908894777298, -0.022440539672970772, 0.020408382639288902, -0.02054072543978691, -0.09643825143575668, -0.04322342947125435, -0.044961269944906235, -0.05960092693567276, 0.041636064648628235, 0.0011985463788732886, 0.0354...
630
630
['Sercan O. Arik', 'Mike Chrzanowski', 'Adam Coates', 'Gregory Diamos', 'Andrew Gibiansky', 'Yongguo Kang', 'Xian Li', 'John Miller', 'Andrew Ng', 'Jonathan Raiman', 'Shubho Sengupta', 'Mohammad Shoeybi']
1702.07825v2
We present Deep Voice, a production-quality text-to-speech system constructed entirely from deep neural networks. Deep Voice lays the groundwork for truly end-to-end neural speech synthesis. The system comprises five major building blocks: a segmentation model for locating phoneme boundaries, a grapheme-to-phoneme conv...
Deep Voice: Real-time Neural Text-to-Speech
2,017
http://arxiv.org/pdf/1702.07825v2
Title Deep Voice Realtime Neural TexttoSpeech Summary present Deep Voice productionquality texttospeech system constructed entirely deep neural network Deep Voice lay groundwork truly endtoend neural speech synthesis system comprises five major building block segmentation model locating phoneme boundary graphemetophone...
[-0.009917370043694973, 0.047601792961359024, 0.015423452481627464, 0.02271011844277382, 0.007447405252605677, -0.04852304980158806, 0.039579831063747406, 0.010334247723221779, -0.051909007132053375, -0.01405063085258007, -0.04730913043022156, -0.025458768010139465, 0.044778551906347275, 0.0770290419459343, 0.020705791...
631
631
['Zichao Yang', 'Zhiting Hu', 'Ruslan Salakhutdinov', 'Taylor Berg-Kirkpatrick']
1702.08139v2
Recent work on generative modeling of text has found that variational auto-encoders (VAE) incorporating LSTM decoders perform worse than simpler LSTM language models (Bowman et al., 2015). This negative result is so far poorly understood, but has been attributed to the propensity of LSTM decoders to ignore conditioning...
Improved Variational Autoencoders for Text Modeling using Dilated Convolutions
2,017
http://arxiv.org/pdf/1702.08139v2
Title Improved Variational Autoencoders Text Modeling using Dilated Convolutions Summary Recent work generative modeling text found variational autoencoders VAE incorporating LSTM decoder perform worse simpler LSTM language model Bowman et al 2015 negative result far poorly understood attributed propensity LSTM decoder...
[0.031139275059103966, 0.08634331822395325, -0.017654934898018837, 0.04814178869128227, -0.019941547885537148, 0.008397197350859642, 0.024982603266835213, -0.0009738968801684678, -0.06876417994499207, -0.0027986273635178804, -0.01723877340555191, -0.04735134541988373, -0.016995685175061226, 0.08497116714715958, 0.02997...
632
632
['Hairong Liu', 'Zhenyao Zhu', 'Xiangang Li', 'Sanjeev Satheesh']
1703.00096v2
Most existing sequence labelling models rely on a fixed decomposition of a target sequence into a sequence of basic units. These methods suffer from two major drawbacks: 1) the set of basic units is fixed, such as the set of words, characters or phonemes in speech recognition, and 2) the decomposition of target sequenc...
Gram-CTC: Automatic Unit Selection and Target Decomposition for Sequence Labelling
2,017
http://arxiv.org/pdf/1703.00096v2
Title GramCTC Automatic Unit Selection Target Decomposition Sequence Labelling Summary existing sequence labelling model rely fixed decomposition target sequence sequence basic unit method suffer two major drawback 1 set basic unit fixed set word character phoneme speech recognition 2 decomposition target sequence fixe...
[0.029830938205122948, 0.04454585537314415, 0.034033939242362976, 0.027482159435749054, -0.016133543103933334, 0.0429430790245533, 0.019833918660879135, 0.0166368018835783, -0.020888321101665497, -0.02717079408466816, -0.000637555553112179, -0.029659390449523926, 0.0503210611641407, 0.0027869197074323893, -0.0121830888...
633
633
['Govardana Sachithanandam Ramachandran', 'Ajay Sohmshetty']
1703.03939v1
We examine Memory Networks for the task of question answering (QA), under common real world scenario where training examples are scarce and under weakly supervised scenario, that is only extrinsic labels are available for training. We propose extensions for the Dynamic Memory Network (DMN), specifically within the atte...
Ask Me Even More: Dynamic Memory Tensor Networks (Extended Model)
2,017
http://arxiv.org/pdf/1703.03939v1
Title Ask Even Dynamic Memory Tensor Networks Extended Model Summary examine Memory Networks task question answering QA common real world scenario training example scarce weakly supervised scenario extrinsic label available training propose extension Dynamic Memory Network DMN specifically within attention mechanism ca...
[0.01754993386566639, -0.01755429059267044, -0.037438519299030304, 0.00472089322283864, -0.007093632128089666, 0.03203025832772255, 0.011884788051247597, -0.020065026357769966, -0.010369201190769672, -0.016145899891853333, -0.0012571010738611221, -0.04139208793640137, -0.013332019560039043, 0.0303955115377903, 0.047639...
634
634
['Lior Fritz', 'David Burshtein']
1703.10356v2
A simplified speech recognition system that uses the maximum mutual information (MMI) criterion is considered. End-to-end training using gradient descent is suggested, similarly to the training of connectionist temporal classification (CTC). We use an MMI criterion with a simple language model in the training stage, an...
Simplified End-to-End MMI Training and Voting for ASR
2,017
http://arxiv.org/pdf/1703.10356v2
Title Simplified EndtoEnd MMI Training Voting ASR Summary simplified speech recognition system us maximum mutual information MMI criterion considered Endtoend training using gradient descent suggested similarly training connectionist temporal classification CTC use MMI criterion simple language model training stage sta...
[-0.026127465069293976, 0.040844596922397614, 0.02768903411924839, 0.027729712426662445, 0.0038226251490414143, 0.022259080782532692, 0.006353217642754316, -0.0174814872443676, -0.0220170971006155, -0.013572930358350277, -0.042083434760570526, -0.01985441893339157, 0.07458272576332092, 0.010344461537897587, -0.01860390...
635
635
['Alec Radford', 'Rafal Jozefowicz', 'Ilya Sutskever']
1704.01444v2
We explore the properties of byte-level recurrent language models. When given sufficient amounts of capacity, training data, and compute time, the representations learned by these models include disentangled features corresponding to high-level concepts. Specifically, we find a single unit which performs sentiment anal...
Learning to Generate Reviews and Discovering Sentiment
2,017
http://arxiv.org/pdf/1704.01444v2
Title Learning Generate Reviews Discovering Sentiment Summary explore property bytelevel recurrent language model given sufficient amount capacity training data compute time representation learned model include disentangled feature corresponding highlevel concept Specifically find single unit performs sentiment analysi...
[0.04611093923449516, 0.05558440461754799, -0.005580800119787455, 0.02697698585689068, -0.04319903999567032, 0.011704379692673683, -0.022924331948161125, 0.0033520604483783245, -0.012462098151445389, -0.06253533810377121, 0.004261876456439495, -0.01009831391274929, -0.013715391047298908, 0.09788276255130768, -0.0238393...
636
636
['Marek Rei']
1704.07156v1
We propose a sequence labeling framework with a secondary training objective, learning to predict surrounding words for every word in the dataset. This language modeling objective incentivises the system to learn general-purpose patterns of semantic and syntactic composition, which are also useful for improving accurac...
Semi-supervised Multitask Learning for Sequence Labeling
2,017
http://arxiv.org/pdf/1704.07156v1
Title Semisupervised Multitask Learning Sequence Labeling Summary propose sequence labeling framework secondary training objective learning predict surrounding word every word dataset language modeling objective incentivises system learn generalpurpose pattern semantic syntactic composition also useful improving accura...
[0.05621160566806793, -0.00344303366728127, 0.021400338038802147, 0.00940660759806633, -0.03424806892871857, 0.039948977530002594, -0.0034205468837171793, -0.040986593812704086, 0.03295162692666054, -0.06236645206809044, -0.036111000925302505, -0.030893437564373016, -0.014165043830871582, 0.058400969952344894, 0.019437...
637
637
['Danhao Zhu', 'Si Shen', 'Xin-Yu Dai', 'Jiajun Chen']
1705.01346v1
Recurrent Neural Network (RNN) has been widely applied for sequence modeling. In RNN, the hidden states at current step are full connected to those at previous step, thus the influence from less related features at previous step may potentially decrease model's learning ability. We propose a simple technique called par...
Going Wider: Recurrent Neural Network With Parallel Cells
2,017
http://arxiv.org/pdf/1705.01346v1
Title Going Wider Recurrent Neural Network Parallel Cells Summary Recurrent Neural Network RNN widely applied sequence modeling RNN hidden state current step full connected previous step thus influence le related feature previous step may potentially decrease model learning ability propose simple technique called paral...
[0.009379123337566853, 0.025332294404506683, -0.024564344435930252, 0.029763156548142433, -0.04892450571060181, 0.014327235519886017, 0.052719030529260635, -0.021398676559329033, -0.05949936434626579, -0.007525046821683645, 0.010927753522992134, -0.06231151521205902, 0.04505552724003792, 0.01007845439016819, 0.02437165...
638
638
['Zhiyuan Tang', 'Dong Wang', 'Yixiang Chen', 'Lantian Li', 'Andrew Abel']
1705.03151v3
Deep neural models, particularly the LSTM-RNN model, have shown great potential for language identification (LID). However, the use of phonetic information has been largely overlooked by most existing neural LID methods, although this information has been used very successfully in conventional phonetic LID systems. We ...
Phonetic Temporal Neural Model for Language Identification
2,017
http://arxiv.org/pdf/1705.03151v3
Title Phonetic Temporal Neural Model Language Identification Summary Deep neural model particularly LSTMRNN model shown great potential language identification LID However use phonetic information largely overlooked existing neural LID method although information used successfully conventional phonetic LID system prese...
[0.010538937523961067, 0.051746804267168045, 0.009194403886795044, 0.02748473547399044, -0.019157063215970993, 0.004291570279747248, 0.04313242807984352, -0.024848293513059616, -0.01615096442401409, 0.013568012043833733, -0.02703378163278103, -0.058014191687107086, 0.07893027365207672, 9.135649634117726e-06, 0.01781799...
639
639
['Hamed Zamani', 'W. Bruce Croft']
1705.03556v2
Learning a high-dimensional dense representation for vocabulary terms, also known as a word embedding, has recently attracted much attention in natural language processing and information retrieval tasks. The embedding vectors are typically learned based on term proximity in a large corpus. This means that the objectiv...
Relevance-based Word Embedding
2,017
http://arxiv.org/pdf/1705.03556v2
Title Relevancebased Word Embedding Summary Learning highdimensional dense representation vocabulary term also known word embedding recently attracted much attention natural language processing information retrieval task embedding vector typically learned based term proximity large corpus mean objective wellknown word ...
[0.055417437106370926, 0.007513559889048338, 0.005542187485843897, 0.029102232307195663, -0.034605324268341064, -0.012237999588251114, -0.03458545729517937, 0.024213548749685287, -0.010251839645206928, -0.06271302700042725, -0.033824384212493896, 0.0020126882009208202, -0.013109618797898293, 0.021384557709097862, 0.006...
640
640
['Tao Lei', 'Wengong Jin', 'Regina Barzilay', 'Tommi Jaakkola']
1705.09037v3
The design of neural architectures for structured objects is typically guided by experimental insights rather than a formal process. In this work, we appeal to kernels over combinatorial structures, such as sequences and graphs, to derive appropriate neural operations. We introduce a class of deep recurrent neural oper...
Deriving Neural Architectures from Sequence and Graph Kernels
2,017
http://arxiv.org/pdf/1705.09037v3
Title Deriving Neural Architectures Sequence Graph Kernels Summary design neural architecture structured object typically guided experimental insight rather formal process work appeal kernel combinatorial structure sequence graph derive appropriate neural operation introduce class deep recurrent neural operation formal...
[0.014733297750353813, 0.04007579758763313, -0.02269957773387432, 0.011624796316027641, -0.007425493560731411, -0.0345165878534317, 0.004953464493155479, 0.03830841928720474, 0.024719465523958206, 0.001643794821575284, 0.05051158741116524, -0.012354633770883083, -0.002367529785260558, 0.09574262797832489, 0.03041026182...
641
641
['Michał Zapotoczny', 'Paweł Rychlikowski', 'Jan Chorowski']
1705.10209v1
We show that a recently proposed neural dependency parser can be improved by joint training on multiple languages from the same family. The parser is implemented as a deep neural network whose only input is orthographic representations of words. In order to successfully parse, the network has to discover how linguistic...
On Multilingual Training of Neural Dependency Parsers
2,017
http://arxiv.org/pdf/1705.10209v1
Title Multilingual Training Neural Dependency Parsers Summary show recently proposed neural dependency parser improved joint training multiple language family parser implemented deep neural network whose input orthographic representation word order successfully parse network discover linguistically relevant concept inf...
[0.01944381557404995, 0.052187707275152206, 0.013219852931797504, 0.08311734348535538, -0.046065762639045715, 0.02014671266078949, 0.0015334389172494411, -0.010520074516534805, -0.010884782299399376, -0.013641287572681904, 0.007354326080530882, -0.037044864147901535, 0.016053467988967896, -0.0013310664799064398, 0.0089...
642
642
['Marian Tietz', 'Tayfun Alpay', 'Johannes Twiefel', 'Stefan Wermter']
1706.02124v2
Ladder networks are a notable new concept in the field of semi-supervised learning by showing state-of-the-art results in image recognition tasks while being compatible with many existing neural architectures. We present the recurrent ladder network, a novel modification of the ladder network, for semi-supervised learn...
Semi-Supervised Phoneme Recognition with Recurrent Ladder Networks
2,017
http://arxiv.org/pdf/1706.02124v2
Title SemiSupervised Phoneme Recognition Recurrent Ladder Networks Summary Ladder network notable new concept field semisupervised learning showing stateoftheart result image recognition task compatible many existing neural architecture present recurrent ladder network novel modification ladder network semisupervised l...
[0.008167542517185211, 0.0533309131860733, 0.02403194271028042, 0.020374435931444168, 0.018757058307528496, -0.005465107969939709, 0.037153396755456924, -0.020483996719121933, 0.0333324559032917, 0.033537209033966064, -0.05731246620416641, -0.034057050943374634, 0.019655730575323105, 0.0464630089700222, 0.0181184187531...
643
643
['Junbo', 'Zhao', 'Yoon Kim', 'Kelly Zhang', 'Alexander M. Rush', 'Yann LeCun']
1706.04223v2
While autoencoders are a key technique in representation learning for continuous structures, such as images or wave forms, developing general-purpose autoencoders for discrete structures, such as text sequence or discretized images, has proven to be more challenging. In particular, discrete inputs make it more difficul...
Adversarially Regularized Autoencoders
2,017
http://arxiv.org/pdf/1706.04223v2
Title Adversarially Regularized Autoencoders Summary autoencoders key technique representation learning continuous structure image wave form developing generalpurpose autoencoders discrete structure text sequence discretized image proven challenging particular discrete input make difficult learn smooth encoder preserve...
[0.003998361993581057, 0.10605769604444504, -0.023995626717805862, 0.047150611877441406, 0.011567512527108192, -0.005254093091934919, 0.02776133269071579, 0.010005134157836437, -0.03373934328556061, 0.008464936167001724, -0.037195928394794464, 0.001652729930356145, -0.002006764989346266, 0.05625103786587715, 0.06576611...
644
644
['Marek Rei', 'Helen Yannakoudakis']
1707.05227v1
We investigate the utility of different auxiliary objectives and training strategies within a neural sequence labeling approach to error detection in learner writing. Auxiliary costs provide the model with additional linguistic information, allowing it to learn general-purpose compositional features that can then be ex...
Auxiliary Objectives for Neural Error Detection Models
2,017
http://arxiv.org/pdf/1707.05227v1
Title Auxiliary Objectives Neural Error Detection Models Summary investigate utility different auxiliary objective training strategy within neural sequence labeling approach error detection learner writing Auxiliary cost provide model additional linguistic information allowing learn generalpurpose compositional feature...
[0.053201671689748764, 0.03163161873817444, 0.020694365724921227, 0.008173013105988503, 0.015069123357534409, 0.03954949602484703, 0.004487816710025072, 0.004229953978210688, -0.010595637373626232, -0.049160826951265335, 0.0034147666301578283, -0.036453258246183395, 0.029932990670204163, 0.0006174460286274552, 0.005157...
645
645
['Youmna Farag', 'Marek Rei', 'Ted Briscoe']
1707.06841v1
We propose a novel word embedding pre-training approach that exploits writing errors in learners' scripts. We compare our method to previous models that tune the embeddings based on script scores and the discrimination between correct and corrupt word contexts in addition to the generic commonly-used embeddings pre-tra...
An Error-Oriented Approach to Word Embedding Pre-Training
2,017
http://arxiv.org/pdf/1707.06841v1
Title ErrorOriented Approach Word Embedding PreTraining Summary propose novel word embedding pretraining approach exploit writing error learner script compare method previous model tune embeddings based script score discrimination correct corrupt word context addition generic commonlyused embeddings pretrained large co...
[0.009240753017365932, 0.028996504843235016, 0.02859247289597988, 0.041179753839969635, 0.01521011721342802, 0.008262046612799168, -0.022857774049043655, 0.0456257164478302, 0.02617158554494381, -0.0855959951877594, -0.03694905713200569, -0.012926281429827213, 0.02177080698311329, 0.04429576173424721, 0.032448224723339...
646
646
['Kartik Goyal', 'Graham Neubig', 'Chris Dyer', 'Taylor Berg-Kirkpatrick']
1708.00111v2
Beam search is a desirable choice of test-time decoding algorithm for neural sequence models because it potentially avoids search errors made by simpler greedy methods. However, typical cross entropy training procedures for these models do not directly consider the behaviour of the final decoding method. As a result, f...
A Continuous Relaxation of Beam Search for End-to-end Training of Neural Sequence Models
2,017
http://arxiv.org/pdf/1708.00111v2
Title Continuous Relaxation Beam Search Endtoend Training Neural Sequence Models Summary Beam search desirable choice testtime decoding algorithm neural sequence model potentially avoids search error made simpler greedy method However typical cross entropy training procedure model directly consider behaviour final deco...
[0.03741150721907616, 0.030471503734588623, 0.03220384195446968, 0.025864267721772194, 0.0011311826528981328, 0.03016337752342224, -0.016953421756625175, 0.013072589412331581, 0.004709469620138407, 0.01821279339492321, -0.028442755341529846, -0.03348269686102867, 0.00608800770714879, -0.007798672188073397, -0.009676429...
647
647
['Stephen Merity', 'Nitish Shirish Keskar', 'Richard Socher']
1708.02182v1
Recurrent neural networks (RNNs), such as long short-term memory networks (LSTMs), serve as a fundamental building block for many sequence learning tasks, including machine translation, language modeling, and question answering. In this paper, we consider the specific problem of word-level language modeling and investi...
Regularizing and Optimizing LSTM Language Models
2,017
http://arxiv.org/pdf/1708.02182v1
Title Regularizing Optimizing LSTM Language Models Summary Recurrent neural network RNNs long shortterm memory network LSTMs serve fundamental building block many sequence learning task including machine translation language modeling question answering paper consider specific problem wordlevel language modeling investi...
[0.03623424842953682, 0.055790528655052185, 0.004418463911861181, 0.06507393717765808, -0.03167427331209183, -0.003000459400936961, 0.013029449619352818, 0.016414618119597435, -0.02476254291832447, -0.019145308062434196, -0.03168972209095955, -0.07381616532802582, 0.05584363266825676, 0.011357031762599945, 0.0064641078...
648
648
['DeLiang Wang', 'Jitong Chen']
1708.07524v1
Speech separation is the task of separating target speech from background interference. Traditionally, speech separation is studied as a signal processing problem. A more recent approach formulates speech separation as a supervised learning problem, where the discriminative patterns of speech, speakers, and background ...
Supervised Speech Separation Based on Deep Learning: An Overview
2,017
http://arxiv.org/pdf/1708.07524v1
Title Supervised Speech Separation Based Deep Learning Overview Summary Speech separation task separating target speech background interference Traditionally speech separation studied signal processing problem recent approach formulates speech separation supervised learning problem discriminative pattern speech speaker...
[0.038107819855213165, 0.01861475221812725, 0.027826201170682907, 0.046899646520614624, 0.004771906416863203, -0.03258495032787323, 0.07156315445899963, -0.01173003576695919, -0.0398559495806694, -0.00904389750212431, -0.06965471059083939, -0.01244617160409689, 0.0264066644012928, 0.004790692590177059, -0.0230887718498...
649
649
['Marek Rei', 'Luana Bulat', 'Douwe Kiela', 'Ekaterina Shutova']
1709.00575v1
The ubiquity of metaphor in our everyday communication makes it an important problem for natural language understanding. Yet, the majority of metaphor processing systems to date rely on hand-engineered features and there is still no consensus in the field as to which features are optimal for this task. In this paper, w...
Grasping the Finer Point: A Supervised Similarity Network for Metaphor Detection
2,017
http://arxiv.org/pdf/1709.00575v1
Title Grasping Finer Point Supervised Similarity Network Metaphor Detection Summary ubiquity metaphor everyday communication make important problem natural language understanding Yet majority metaphor processing system date rely handengineered feature still consensus field feature optimal task paper present first deep ...
[0.05495205149054527, 0.00798508059233427, -0.023405704647302628, 0.03188026323914528, -0.017655406147241592, 0.023375147953629494, -0.003949107602238655, -0.014763303101062775, -0.01064575556665659, -0.025932954624295235, 0.009613321162760258, -0.0046531278640031815, 0.022552382200956345, 0.010603792034089565, 0.01836...
650
650
['Danushka Bollegala', 'Kohei Hayashi', 'Ken-ichi Kawarabayashi']
1709.06671v1
Distributed word embeddings have shown superior performances in numerous Natural Language Processing (NLP) tasks. However, their performances vary significantly across different tasks, implying that the word embeddings learnt by those methods capture complementary aspects of lexical semantics. Therefore, we believe tha...
Think Globally, Embed Locally --- Locally Linear Meta-embedding of Words
2,017
http://arxiv.org/pdf/1709.06671v1
Title Think Globally Embed Locally Locally Linear Metaembedding Words Summary Distributed word embeddings shown superior performance numerous Natural Language Processing NLP task However performance vary significantly across different task implying word embeddings learnt method capture complementary aspect lexical sema...
[0.012094358913600445, 0.01215886790305376, 0.007214866112917662, 0.071959488093853, -0.008110486902296543, -0.003673032158985734, 0.007556939963251352, -0.031710900366306305, -0.05535687133669853, -0.05274314433336258, -0.04593686759471893, 0.07059600949287415, 0.0120661910623312, -0.016054438427090645, 0.029526535421...
651
651
['Bin Bi', 'Hao Ma']
1709.09749v1
Previous studies have demonstrated the empirical success of word embeddings in various applications. In this paper, we investigate the problem of learning distributed representations for text documents which many machine learning algorithms take as input for a number of NLP tasks. We propose a neural network model, K...
KeyVec: Key-semantics Preserving Document Representations
2,017
http://arxiv.org/pdf/1709.09749v1
Title KeyVec Keysemantics Preserving Document Representations Summary Previous study demonstrated empirical success word embeddings various application paper investigate problem learning distributed representation text document many machine learning algorithm take input number NLP task propose neural network model KeyV...
[0.04206375405192375, -0.0009373161592520773, 0.021352212876081467, 0.03221099078655243, -0.022413088008761406, 0.015619372017681599, -0.0132781146094203, 0.039899419993162155, -0.010965495370328426, -0.0836675688624382, 0.008697493001818657, 0.010289817117154598, -0.008964930661022663, 0.0367455892264843, 0.0169996228...
652
652
['Shuai Tang', 'Hailin Jin', 'Chen Fang', 'Zhaowen Wang', 'Virginia R. de Sa']
1710.10380v2
Context information plays an important role in human language understanding, and it is also useful for machines to learn vector representations of language. In this paper, we explore an asymmetric encoder-decoder structure for unsupervised context-based sentence representation learning. As a result, we build an encoder...
Exploring Asymmetric Encoder-Decoder Structure for Context-based Sentence Representation Learning
2,017
http://arxiv.org/pdf/1710.10380v2
Title Exploring Asymmetric EncoderDecoder Structure Contextbased Sentence Representation Learning Summary Context information play important role human language understanding also useful machine learn vector representation language paper explore asymmetric encoderdecoder structure unsupervised contextbased sentence rep...
[0.031040281057357788, 0.012019283138215542, -0.020153719931840897, 0.10539435595273972, -0.011638340540230274, 0.0379156768321991, 0.016123764216899872, -0.021891169250011444, -0.071658194065094, -0.08328808844089508, -0.021733807399868965, -0.017984628677368164, -0.016857003793120384, 0.02559058926999569, -0.00907099...
653
653
['Qiming Chen', 'Ren Wu']
1712.09662v1
The Convolution Neural Network (CNN) has demonstrated the unique advantage in audio, image and text learning; recently it has also challenged Recurrent Neural Networks (RNNs) with long short-term memory cells (LSTM) in sequence-to-sequence learning, since the computations involved in CNN are easily parallelizable where...
CNN Is All You Need
2,017
http://arxiv.org/pdf/1712.09662v1
Title CNN Need Summary Convolution Neural Network CNN demonstrated unique advantage audio image text learning recently also challenged Recurrent Neural Networks RNNs long shortterm memory cell LSTM sequencetosequence learning since computation involved CNN easily parallelizable whereas involved RNN mostly sequential le...
[0.023207545280456543, 0.055911850184202194, -0.016193347051739693, 0.03825205937027931, 0.010565239936113358, -0.011192799545824528, 0.031835008412599564, 0.025277020409703255, -0.055737681686878204, -0.008467582054436207, -0.018891112878918648, -0.024584462866187096, 0.026755500584840775, -0.0033087453339248896, 0.01...
654
654
['Tim Rocktäschel']
1712.09687v1
The current state-of-the-art in many natural language processing and automated knowledge base completion tasks is held by representation learning methods which learn distributed vector representations of symbols via gradient-based optimization. They require little or no hand-crafted features, thus avoiding the need for...
Combining Representation Learning with Logic for Language Processing
2,017
http://arxiv.org/pdf/1712.09687v1
Title Combining Representation Learning Logic Language Processing Summary current stateoftheart many natural language processing automated knowledge base completion task held representation learning method learn distributed vector representation symbol via gradientbased optimization require little handcrafted feature t...
[0.04245384782552719, -0.010935433208942413, -0.00979213509708643, 0.019917266443371773, -0.03041921928524971, 0.02238231897354126, -0.009313097223639488, 0.012592816725373268, 0.06660055369138718, -0.09160167723894119, 0.018299521878361702, 0.04960286617279053, 0.0038200649432837963, 0.05390718951821327, 0.01200169976...
655
655
['Dietmar Volz']
1308.1603v4
It will be shown that according to theorems of K. Menger, every neuron grid if identified with a curve is able to preserve the adopted qualitative structure of a data space. Furthermore, if this identification is made, the neuron grid structure can always be mapped to a subset of a universal neuron grid which is constr...
A Note on Topology Preservation in Classification, and the Construction of a Universal Neuron Grid
2,013
http://arxiv.org/pdf/1308.1603v4
Title Note Topology Preservation Classification Construction Universal Neuron Grid Summary shown according theorem K Menger every neuron grid identified curve able preserve adopted qualitative structure data space Furthermore identification made neuron grid structure always mapped subset universal neuron grid construct...
[-0.018376406282186508, 0.009212622418999672, -0.02799879014492035, 0.002389247063547373, -0.012164520099759102, 0.007206049747765064, 0.07893484085798264, -0.014987609349191189, 0.008515902794897556, 0.027642400935292244, -0.05880561098456383, 0.001487764879129827, 0.009216845966875553, 0.04748561978340149, 0.06948754...
656
656
['Louis Yuanlong Shao']
1210.8442v3
One conjecture in both deep learning and classical connectionist viewpoint is that the biological brain implements certain kinds of deep networks as its back-end. However, to our knowledge, a detailed correspondence has not yet been set up, which is important if we want to bridge between neuroscience and machine learni...
Linear-Nonlinear-Poisson Neuron Networks Perform Bayesian Inference On Boltzmann Machines
2,012
http://arxiv.org/pdf/1210.8442v3
Title LinearNonlinearPoisson Neuron Networks Perform Bayesian Inference Boltzmann Machines Summary One conjecture deep learning classical connectionist viewpoint biological brain implement certain kind deep network backend However knowledge detailed correspondence yet set important want bridge neuroscience machine lear...
[-0.04433208703994751, 0.0404491052031517, -0.029001904651522636, 0.02259504236280918, -0.025148529559373856, 0.007500724866986275, 0.043871693313121796, -0.01455585565418005, -0.03772580251097679, -0.005837182514369488, -0.030823180451989174, 0.01564974896609783, 0.004207680933177471, 0.05848691985011101, 0.0803089365...
657
657
['Enmei Tu', 'Nikola Kasabov', 'Jie Yang']
1603.05594v1
This paper proposes a new method for an optimized mapping of temporal variables, describing a temporal stream data, into the recently proposed NeuCube spiking neural network architecture. This optimized mapping extends the use of the NeuCube, which was initially designed for spatiotemporal brain data, to work on arbitr...
Mapping Temporal Variables into the NeuCube for Improved Pattern Recognition, Predictive Modelling and Understanding of Stream Data
2,016
http://arxiv.org/pdf/1603.05594v1
Title Mapping Temporal Variables NeuCube Improved Pattern Recognition Predictive Modelling Understanding Stream Data Summary paper proposes new method optimized mapping temporal variable describing temporal stream data recently proposed NeuCube spiking neural network architecture optimized mapping extends use NeuCube i...
[-0.03617202118039131, -0.012301030568778515, -0.032941706478595734, -0.005760688800364733, 0.004749706480652094, 0.016142960637807846, 0.019739855080842972, -0.02388072945177555, -0.021417731419205666, 0.01587310992181301, 0.03027486428618431, -0.020441768690943718, 0.025372151285409927, 0.1118316650390625, 0.02946531...
658
658
['Jörn Hees', 'Rouven Bauer', 'Joachim Folz', 'Damian Borth', 'Andreas Dengel']
1607.07249v3
Efficient usage of the knowledge provided by the Linked Data community is often hindered by the need for domain experts to formulate the right SPARQL queries to answer questions. For new questions they have to decide which datasets are suitable and in which terminology and modelling style to phrase the SPARQL query. ...
An Evolutionary Algorithm to Learn SPARQL Queries for Source-Target-Pairs: Finding Patterns for Human Associations in DBpedia
2,016
http://arxiv.org/pdf/1607.07249v3
Title Evolutionary Algorithm Learn SPARQL Queries SourceTargetPairs Finding Patterns Human Associations DBpedia Summary Efficient usage knowledge provided Linked Data community often hindered need domain expert formulate right SPARQL query answer question new question decide datasets suitable terminology modelling styl...
[0.05303521826863289, 0.0714755430817604, -0.023572899401187897, -0.0002667703665792942, 0.0017019326332956553, -0.007072402164340019, 0.00022472212731372565, 0.00924236886203289, 0.014618105255067348, -0.013973248191177845, 0.010560072027146816, 0.06168920919299126, -0.004193740896880627, 0.08381201326847076, -0.00691...
659
659
['Anthony L. Caterini', 'Dong Eui Chang']
1608.04374v2
In this paper, a geometric framework for neural networks is proposed. This framework uses the inner product space structure underlying the parameter set to perform gradient descent not in a component-based form, but in a coordinate-free manner. Convolutional neural networks are described in this framework in a compact ...
A Geometric Framework for Convolutional Neural Networks
2,016
http://arxiv.org/pdf/1608.04374v2
Title Geometric Framework Convolutional Neural Networks Summary paper geometric framework neural network proposed framework us inner product space structure underlying parameter set perform gradient descent componentbased form coordinatefree manner Convolutional neural network described framework compact form gradient ...
[-0.0062927547842264175, 0.02175666019320488, -0.01843675784766674, 0.05009322240948677, 0.00307169440202415, 0.0009931556414812803, 0.06940880417823792, -0.007466486655175686, -0.00551449041813612, 0.02989242784678936, 0.01591523550450802, 0.00019580336811486632, 0.0005112610524520278, 0.043925248086452484, 0.03991241...
660
660
['Anthony Caterini', 'Dong Eui Chang']
1610.01549v2
Deep Neural Networks (DNNs) have become very popular for prediction in many areas. Their strength is in representation with a high number of parameters that are commonly learned via gradient descent or similar optimization methods. However, the representation is non-standardized, and the gradient calculation methods ar...
A Novel Representation of Neural Networks
2,016
http://arxiv.org/pdf/1610.01549v2
Title Novel Representation Neural Networks Summary Deep Neural Networks DNNs become popular prediction many area strength representation high number parameter commonly learned via gradient descent similar optimization method However representation nonstandardized gradient calculation method often performed using compon...
[-0.04897966608405113, 0.018185332417488098, -0.04835548624396324, 0.03892781585454941, 0.01695767045021057, -0.028132939711213112, 0.06393872946500778, -0.034992609173059464, -0.01274050772190094, -0.003143599722534418, -0.04479924961924553, -0.010154358111321926, 0.009778140112757683, 0.01962733082473278, 0.026690669...
661
661
['Ardavan Salehi Nobandegani', 'Thomas R. Shultz']
1701.05004v1
Humans are not only adept in recognizing what class an input instance belongs to (i.e., classification task), but perhaps more remarkably, they can imagine (i.e., generate) plausible instances of a desired class with ease, when prompted. Inspired by this, we propose a framework which allows transforming Cascade-Correla...
Converting Cascade-Correlation Neural Nets into Probabilistic Generative Models
2,017
http://arxiv.org/pdf/1701.05004v1
Title Converting CascadeCorrelation Neural Nets Probabilistic Generative Models Summary Humans adept recognizing class input instance belongs ie classification task perhaps remarkably imagine ie generate plausible instance desired class ease prompted Inspired propose framework allows transforming CascadeCorrelation Neu...
[0.00014280565665103495, 0.09656462073326111, -0.032682426273822784, 0.012405757792294025, -0.03379938006401062, 0.002445181366056204, 0.053696729242801666, -0.02378443256020546, -0.009187174960970879, 0.014752513729035854, -0.0019390084780752659, 0.034141041338443756, 0.009422731585800648, 0.023686226457357407, -0.011...
662
662
['Ludvig Ericson', 'Rendani Mbuvha']
1701.05130v1
Artificial Neural Networks (ANNs) have received increasing attention in recent years with applications that span a wide range of disciplines including vital domains such as medicine, network security and autonomous transportation. However, neural network architectures are becoming increasingly complex and with an incre...
On the Performance of Network Parallel Training in Artificial Neural Networks
2,017
http://arxiv.org/pdf/1701.05130v1
Title Performance Network Parallel Training Artificial Neural Networks Summary Artificial Neural Networks ANNs received increasing attention recent year application span wide range discipline including vital domain medicine network security autonomous transportation However neural network architecture becoming increasi...
[0.00465263519436121, -0.006094884127378464, -0.03394026309251785, 0.025037458166480064, 0.006173739209771156, 0.01080020610243082, 0.08450014889240265, -0.028840018436312675, -0.03439641743898392, -0.01235903799533844, -0.002023539040237665, 0.02701936475932598, 0.008542477153241634, -0.019386006519198418, 0.023082798...
663
663
['Misha Denil', 'Sergio Gómez Colmenarejo', 'Serkan Cabi', 'David Saxton', 'Nando de Freitas']
1706.06383v1
We build deep RL agents that execute declarative programs expressed in formal language. The agents learn to ground the terms in this language in their environment, and can generalize their behavior at test time to execute new programs that refer to objects that were not referenced during training. The agents develop di...
Programmable Agents
2,017
http://arxiv.org/pdf/1706.06383v1
Title Programmable Agents Summary build deep RL agent execute declarative program expressed formal language agent learn ground term language environment generalize behavior test time execute new program refer object referenced training agent develop disentangled interpretable representation allow generalize wide variet...
[0.015608362853527069, 0.04042176902294159, -0.008841241709887981, 0.021490028128027916, -0.04310496523976326, 0.0172731876373291, 0.004734480753540993, -0.02396218851208687, -0.006850042380392551, -0.029864156618714333, 0.027934923768043518, 0.06956136226654053, -0.03755027800798416, 0.12334099411964417, -0.0050628301...
664
664
['Wojciech Samek', 'Thomas Wiegand', 'Klaus-Robert Müller']
1708.08296v1
With the availability of large databases and recent improvements in deep learning methodology, the performance of AI systems is reaching or even exceeding the human level on an increasing number of complex tasks. Impressive examples of this development can be found in domains such as image classification, sentiment ana...
Explainable Artificial Intelligence: Understanding, Visualizing and Interpreting Deep Learning Models
2,017
http://arxiv.org/pdf/1708.08296v1
Title Explainable Artificial Intelligence Understanding Visualizing Interpreting Deep Learning Models Summary availability large database recent improvement deep learning methodology performance AI system reaching even exceeding human level increasing number complex task Impressive example development found domain imag...
[0.004354760050773621, 0.025785479694604874, -0.040497828274965286, 0.008426116779446602, 0.0018225862877443433, 0.011091730557382107, 0.017602263018488884, 0.015021657571196556, -0.03170483559370041, 0.016350071877241135, 0.003045437391847372, 0.02138860709965229, -0.0071899862959980965, 0.0925000011920929, 0.01675647...
665
665
['Maithra Raghu', 'Alex Irpan', 'Jacob Andreas', 'Robert Kleinberg', 'Quoc V. Le', 'Jon Kleinberg']
1711.02301v3
Deep reinforcement learning has achieved many recent successes, but our understanding of its strengths and limitations is hampered by the lack of rich environments in which we can fully characterize optimal behavior, and correspondingly diagnose individual actions against such a characterization. Here we consider a fam...
Can Deep Reinforcement Learning Solve Erdos-Selfridge-Spencer Games?
2,017
http://arxiv.org/pdf/1711.02301v3
Title Deep Reinforcement Learning Solve ErdosSelfridgeSpencer Games Summary Deep reinforcement learning achieved many recent success understanding strength limitation hampered lack rich environment fully characterize optimal behavior correspondingly diagnose individual action characterization consider family combinator...
[0.00046241149539127946, 0.012712793424725533, -0.039370663464069366, 0.024848148226737976, -0.035569336265325546, -0.009846172295510769, 0.011720039881765842, 0.0038367644883692265, -0.031720489263534546, 0.046587537974119186, 0.00973449181765318, 0.033039599657058716, -0.055030446499586105, 0.05365367978811264, 0.004...
666
666
['Christopher J. Cueva', 'Xue-Xin Wei']
1803.07770v1
Decades of research on the neural code underlying spatial navigation have revealed a diverse set of neural response properties. The Entorhinal Cortex (EC) of the mammalian brain contains a rich set of spatial correlates, including grid cells which encode space using tessellating patterns. However, the mechanisms and fu...
Emergence of grid-like representations by training recurrent neural networks to perform spatial localization
2,018
http://arxiv.org/pdf/1803.07770v1
Title Emergence gridlike representation training recurrent neural network perform spatial localization Summary Decades research neural code underlying spatial navigation revealed diverse set neural response property Entorhinal Cortex EC mammalian brain contains rich set spatial correlate including grid cell encode spac...
[0.0009491996024735272, 0.018944118171930313, -0.03762098774313927, -0.005847285035997629, 0.01585310883820057, 0.039605412632226944, 0.015460417605936527, -0.018404824659228325, -0.020391466096043587, 0.03385598212480545, -0.029813308268785477, -0.029472600668668747, 0.022981125861406326, 0.05065745860338211, 0.083313...
667
667
['Dasika Ratna Deepthi', 'Sujeet Kuchibhotla', 'K. Eswaran']
0712.0932v1
In this paper, we present a Mirroring Neural Network architecture to perform non-linear dimensionality reduction and Object Recognition using a reduced lowdimensional characteristic vector. In addition to dimensionality reduction, the network also reconstructs (mirrors) the original high-dimensional input vector from t...
Dimensionality Reduction and Reconstruction using Mirroring Neural Networks and Object Recognition based on Reduced Dimension Characteristic Vector
2,007
http://arxiv.org/pdf/0712.0932v1
Title Dimensionality Reduction Reconstruction using Mirroring Neural Networks Object Recognition based Reduced Dimension Characteristic Vector Summary paper present Mirroring Neural Network architecture perform nonlinear dimensionality reduction Object Recognition using reduced lowdimensional characteristic vector addi...
[-0.024565834552049637, 0.008700142614543438, -0.027781834825873375, 0.06486279517412186, -0.012282852083444595, 0.009073599241673946, 0.050775375217199326, 0.015378891490399837, -0.05731024965643883, 0.012595757842063904, -0.0018333224579691887, 0.008624824695289135, 0.029290223494172096, 0.02354268915951252, 0.034534...
668
668
['Yongnan Ji', 'Pierre-Yves Herve', 'Uwe Aickelin', 'Alain Pitiot']
1004.3708v1
Inter-subject parcellation of functional Magnetic Resonance Imaging (fMRI) data based on a standard General Linear Model (GLM)and spectral clustering was recently proposed as a means to alleviate the issues associated with spatial normalization in fMRI. However, for all its appeal, a GLM-based parcellation approach int...
Parcellation of fMRI Datasets with ICA and PLS-A Data Driven Approach
2,010
http://arxiv.org/pdf/1004.3708v1
Title Parcellation fMRI Datasets ICA PLSA Data Driven Approach Summary Intersubject parcellation functional Magnetic Resonance Imaging fMRI data based standard General Linear Model GLMand spectral clustering recently proposed mean alleviate issue associated spatial normalization fMRI However appeal GLMbased parcellatio...
[-0.00980097334831953, 0.022604599595069885, -0.043833956122398376, -0.005876052659004927, 0.012391098774969578, 0.04289724677801132, 0.06515447795391083, 0.04038946330547333, 0.009628823027014732, 0.06108696386218071, -0.06843267381191254, -0.038438066840171814, 0.05879276618361473, 0.02290216088294983, 0.052266221493...
669
669
['N. Popescu-Bodorin', 'V. E. Balas', 'I. M. Motoc']
1110.6483v2
The main topic discussed in this paper is how to use intelligence for biometric decision defuzzification. A neural training model is proposed and tested here as a possible solution for dealing with natural fuzzification that appears between the intra- and inter-class distribution of scores computed during iris recognit...
Iris Codes Classification Using Discriminant and Witness Directions
2,011
http://arxiv.org/pdf/1110.6483v2
Title Iris Codes Classification Using Discriminant Witness Directions Summary main topic discussed paper use intelligence biometric decision defuzzification neural training model proposed tested possible solution dealing natural fuzzification appears intra interclass distribution score computed iris recognition test sh...
[-0.010166543535888195, 0.02445843070745468, -0.03575289249420166, 0.02355828322470188, 0.008509612642228603, 0.006906767375767231, 0.0847771093249321, 0.030097832903265953, 0.028248466551303864, 0.035606518387794495, 0.008046114817261696, -0.011559037491679192, 0.0764433741569519, 0.045154646039009094, -0.002299973974...
670
670
['Harris Georgiou']
0910.3348v1
Medical Informatics and the application of modern signal processing in the assistance of the diagnostic process in medical imaging is one of the more recent and active research areas today. This thesis addresses a variety of issues related to the general problem of medical image analysis, specifically in mammography, a...
Algorithms for Image Analysis and Combination of Pattern Classifiers with Application to Medical Diagnosis
2,009
http://arxiv.org/pdf/0910.3348v1
Title Algorithms Image Analysis Combination Pattern Classifiers Application Medical Diagnosis Summary Medical Informatics application modern signal processing assistance diagnostic process medical imaging one recent active research area today thesis address variety issue related general problem medical image analysis s...
[0.021188486367464066, 0.012472002767026424, -0.03954780101776123, 0.0048455409705638885, -0.020418690517544746, 0.026509681716561317, 0.04262324795126915, 0.07307633012533188, -0.034143172204494476, 0.020297590643167496, 0.10182428359985352, -0.00038924592081457376, 0.05464025214314461, 0.07057800889015198, -0.0151524...
671
671
['Anh Nguyen', 'Jason Yosinski', 'Jeff Clune']
1412.1897v4
Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences rem...
Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images
2,014
http://arxiv.org/pdf/1412.1897v4
Title Deep Neural Networks Easily Fooled High Confidence Predictions Unrecognizable Images Summary Deep neural network DNNs recently achieving stateoftheart performance variety patternrecognition task notably visual classification problem Given DNNs able classify object image nearhumanlevel performance question natural...
[0.010116065852344036, 0.09094808995723724, -0.03562874346971512, 0.05571739003062248, -0.002176515059545636, 0.0061898017302155495, 0.055231500416994095, -0.015633421018719673, -0.032679539173841476, -0.002561143832281232, -0.01845097541809082, 0.036392152309417725, 0.026111774146556854, 0.05467673018574715, 0.0502916...
672
672
['Xinyu Wu', 'Vishal Saxena', 'Kehan Zhu']
1506.01072v2
A neuromorphic chip that combines CMOS analog spiking neurons and memristive synapses offers a promising solution to brain-inspired computing, as it can provide massive neural network parallelism and density. Previous hybrid analog CMOS-memristor approaches required extensive CMOS circuitry for training, and thus elimi...
Homogeneous Spiking Neuromorphic System for Real-World Pattern Recognition
2,015
http://arxiv.org/pdf/1506.01072v2
Title Homogeneous Spiking Neuromorphic System RealWorld Pattern Recognition Summary neuromorphic chip combine CMOS analog spiking neuron memristive synapsis offer promising solution braininspired computing provide massive neural network parallelism density Previous hybrid analog CMOSmemristor approach required extensiv...
[-0.053706616163253784, -0.03694482147693634, -0.052267879247665405, 0.05803336948156357, 0.012484094128012657, 0.00016219884855672717, 0.010011780075728893, -0.00529831787571311, 0.0475807823240757, -0.001519152894616127, -0.042839165776968, 0.001372783095575869, -0.004398516844958067, 0.05477315932512283, 0.031505819...
673
673
['Ven Jyn Kok', 'Mei Kuan Lim', 'Chee Seng Chan']
1511.06586v1
Although the traits emerged in a mass gathering are often non-deliberative, the act of mass impulse may lead to irre- vocable crowd disasters. The two-fold increase of carnage in crowd since the past two decades has spurred significant advances in the field of computer vision, towards effective and proactive crowd surv...
Crowd Behavior Analysis: A Review where Physics meets Biology
2,015
http://arxiv.org/pdf/1511.06586v1
Title Crowd Behavior Analysis Review Physics meet Biology Summary Although trait emerged mass gathering often nondeliberative act mass impulse may lead irre vocable crowd disaster twofold increase carnage crowd since past two decade spurred significant advance field computer vision towards effective proactive crowd sur...
[-0.002358394907787442, 0.015343783423304558, -0.02958426997065544, -0.014622044749557972, -0.010780510492622852, -0.021123582497239113, -0.0009222599328495562, 0.007528827525675297, -0.02057489939033985, 0.004340232815593481, 0.039204876869916916, 0.007756952196359634, -0.018435323610901833, 0.006776188500225544, 0.06...
674
674
['Vlado Menkovski', 'Zharko Aleksovski', 'Axel Saalbach', 'Hannes Nickisch']
1512.05986v1
Convolutional neural networks demonstrated outstanding empirical results in computer vision and speech recognition tasks where labeled training data is abundant. In medical imaging, there is a huge variety of possible imaging modalities and contrasts, where annotated data is usually very scarce. We present two approach...
Can Pretrained Neural Networks Detect Anatomy?
2,015
http://arxiv.org/pdf/1512.05986v1
Title Pretrained Neural Networks Detect Anatomy Summary Convolutional neural network demonstrated outstanding empirical result computer vision speech recognition task labeled training data abundant medical imaging huge variety possible imaging modality contrast annotated data usually scarce present two approach deal ch...
[0.014294211752712727, 0.06194915249943733, -0.008607087656855583, 0.014003215357661247, 0.02963629737496376, 0.034846555441617966, 0.02385333739221096, 0.03210815414786339, 0.0029922490939497948, 0.022602366283535957, -0.048182617872953415, -0.0062265945598483086, 0.025089100003242493, 0.026192206889390945, 0.01063295...
675
675
['L. M. Rasdi Rere', 'Mohamad Ivan Fanany', 'Aniati Murni Arymurthy']
1610.01925v1
A typical modern optimization technique is usually either heuristic or metaheuristic. This technique has managed to solve some optimization problems in the research area of science, engineering, and industry. However, implementation strategy of metaheuristic for accuracy improvement on convolution neural networks (CNN)...
Metaheuristic Algorithms for Convolution Neural Network
2,016
http://arxiv.org/pdf/1610.01925v1
Title Metaheuristic Algorithms Convolution Neural Network Summary typical modern optimization technique usually either heuristic metaheuristic technique managed solve optimization problem research area science engineering industry However implementation strategy metaheuristic accuracy improvement convolution neural net...
[0.02029944397509098, 0.006519587244838476, -0.013802915811538696, 0.011240843683481216, -0.024036340415477753, -0.01870868168771267, 0.038991380482912064, -0.0033244576770812273, -0.07844418287277222, -0.0439758226275444, 0.003446210641413927, 0.030755717307329178, 0.011416098102927208, 0.026213426142930984, 0.0012278...
676
676
['Jin-Hwa Kim', 'Kyoung-Woon On', 'Woosang Lim', 'Jeonghee Kim', 'Jung-Woo Ha', 'Byoung-Tak Zhang']
1610.04325v4
Bilinear models provide rich representations compared with linear models. They have been applied in various visual tasks, such as object recognition, segmentation, and visual question-answering, to get state-of-the-art performances taking advantage of the expanded representations. However, bilinear representations tend...
Hadamard Product for Low-rank Bilinear Pooling
2,016
http://arxiv.org/pdf/1610.04325v4
Title Hadamard Product Lowrank Bilinear Pooling Summary Bilinear model provide rich representation compared linear model applied various visual task object recognition segmentation visual questionanswering get stateoftheart performance taking advantage expanded representation However bilinear representation tend highdi...
[0.0004682496073655784, 0.014315168373286724, -0.02317340485751629, 0.02531825192272663, -0.029307929798960686, 0.017951473593711853, 0.04475909471511841, 0.023417161777615547, 0.02642686665058136, -0.04970598593354225, 0.0156943928450346, -0.0571695975959301, -0.01680913008749485, 0.017787916585803032, 0.0564383789896...
677
677
['Aojun Zhou', 'Anbang Yao', 'Yiwen Guo', 'Lin Xu', 'Yurong Chen']
1702.03044v2
This paper presents incremental network quantization (INQ), a novel method, targeting to efficiently convert any pre-trained full-precision convolutional neural network (CNN) model into a low-precision version whose weights are constrained to be either powers of two or zero. Unlike existing methods which are struggled ...
Incremental Network Quantization: Towards Lossless CNNs with Low-Precision Weights
2,017
http://arxiv.org/pdf/1702.03044v2
Title Incremental Network Quantization Towards Lossless CNNs LowPrecision Weights Summary paper present incremental network quantization INQ novel method targeting efficiently convert pretrained fullprecision convolutional neural network CNN model lowprecision version whose weight constrained either power two zero Unli...
[-0.02954309619963169, 0.0442105196416378, 0.0049345605075359344, 0.053062520921230316, 0.03744306415319443, -0.01931591145694256, 0.034151557832956314, 0.03957200050354004, -0.053639624267816544, 0.000521279638633132, 0.009054725989699364, 0.016831547021865845, -0.01895867846906185, 0.02614227496087551, 0.031029893085...
678
678
['Dhanesh Ramachandram', 'Terrance DeVries']
1703.03372v3
We present a method for skin lesion segmentation for the ISIC 2017 Skin Lesion Segmentation Challenge. Our approach is based on a Fully Convolutional Network architecture which is trained end to end, from scratch, on a limited dataset. Our semantic segmentation architecture utilizes several recent innovations in partic...
LesionSeg: Semantic segmentation of skin lesions using Deep Convolutional Neural Network
2,017
http://arxiv.org/pdf/1703.03372v3
Title LesionSeg Semantic segmentation skin lesion using Deep Convolutional Neural Network Summary present method skin lesion segmentation ISIC 2017 Skin Lesion Segmentation Challenge approach based Fully Convolutional Network architecture trained end end scratch limited dataset semantic segmentation architecture utiliz...
[0.03243638575077057, -0.0013901983620598912, 0.012587620876729488, 0.012158333323895931, 0.03590473160147667, 0.008580729365348816, 0.02405596151947975, 0.016792617738246918, -0.015188582241535187, 0.024354035034775734, 0.021119659766554832, 0.024920549243688583, -0.012242786586284637, 0.035538893193006516, -0.0056605...
679
679
['Priyadarshini Panda', 'Gopalakrishnan Srinivasan', 'Kaushik Roy']
1703.03854v2
Brain-inspired learning models attempt to mimic the cortical architecture and computations performed in the neurons and synapses constituting the human brain to achieve its efficiency in cognitive tasks. In this work, we present convolutional spike timing dependent plasticity based feature learning with biologically pl...
Convolutional Spike Timing Dependent Plasticity based Feature Learning in Spiking Neural Networks
2,017
http://arxiv.org/pdf/1703.03854v2
Title Convolutional Spike Timing Dependent Plasticity based Feature Learning Spiking Neural Networks Summary Braininspired learning model attempt mimic cortical architecture computation performed neuron synapsis constituting human brain achieve efficiency cognitive task work present convolutional spike timing dependent...
[-0.0192340686917305, 0.00753026083111763, -0.027655042707920074, 0.07184738665819168, 0.02239409275352955, -0.022748205810785294, 0.03357100859284401, -0.01638604700565338, 0.025535887107253075, 0.027913639321923256, -0.05634466931223869, 0.04285196214914322, 0.0009294417104683816, 0.09864293038845062, 0.0287422630935...
680
680
['Shumeet Baluja', 'Ian Fischer']
1703.09387v1
Multiple different approaches of generating adversarial examples have been proposed to attack deep neural networks. These approaches involve either directly computing gradients with respect to the image pixels, or directly solving an optimization on the image pixels. In this work, we present a fundamentally new method ...
Adversarial Transformation Networks: Learning to Generate Adversarial Examples
2,017
http://arxiv.org/pdf/1703.09387v1
Title Adversarial Transformation Networks Learning Generate Adversarial Examples Summary Multiple different approach generating adversarial example proposed attack deep neural network approach involve either directly computing gradient respect image pixel directly solving optimization image pixel work present fundament...
[0.02422841638326645, 0.03188668191432953, -0.02534564398229122, 0.036693260073661804, -0.010438768193125725, -0.04462973400950432, 0.057960670441389084, -0.024842381477355957, -0.05617695674300194, -0.027730831876397133, -0.02204558625817299, 0.06027965992689133, 0.011222568340599537, 0.022123096510767937, 0.059825230...
681
681
['Devinder Kumar', 'Graham W Taylor', 'Alexander Wong']
1709.01574v1
Deep learning has been shown to outperform traditional machine learning algorithms across a wide range of problem domains. However, current deep learning algorithms have been criticized as uninterpretable "black-boxes" which cannot explain their decision making processes. This is a major shortcoming that prevents the w...
Opening the Black Box of Financial AI with CLEAR-Trade: A CLass-Enhanced Attentive Response Approach for Explaining and Visualizing Deep Learning-Driven Stock Market Prediction
2,017
http://arxiv.org/pdf/1709.01574v1
Title Opening Black Box Financial AI CLEARTrade CLassEnhanced Attentive Response Approach Explaining Visualizing Deep LearningDriven Stock Market Prediction Summary Deep learning shown outperform traditional machine learning algorithm across wide range problem domain However current deep learning algorithm criticized u...
[-0.040640443563461304, 0.016146590933203697, -0.05602003633975983, -0.03990969434380531, 0.028110221028327942, -0.006837813183665276, 0.04842669889330864, 0.03750529885292053, -0.010637651197612286, 0.01545463316142559, 0.03076239489018917, 0.022501042112708092, -0.0024728206917643547, 0.1408827304840088, 0.0214335024...
682
682
['Mohammad Javad Shafiee', 'Brendan Chywl', 'Francis Li', 'Alexander Wong']
1709.05943v1
Object detection is considered one of the most challenging problems in this field of computer vision, as it involves the combination of object classification and object localization within a scene. Recently, deep neural networks (DNNs) have been demonstrated to achieve superior object detection performance compared to ...
Fast YOLO: A Fast You Only Look Once System for Real-time Embedded Object Detection in Video
2,017
http://arxiv.org/pdf/1709.05943v1
Title Fast YOLO Fast Look System Realtime Embedded Object Detection Video Summary Object detection considered one challenging problem field computer vision involves combination object classification object localization within scene Recently deep neural network DNNs demonstrated achieve superior object detection perform...
[0.014544973149895668, 0.03064371645450592, -0.000905121210962534, 0.0811096802353859, 0.007861118763685226, 0.009515101090073586, 0.03478757292032242, 0.019740769639611244, -0.017672598361968994, -0.0129587696865201, 0.027494484558701515, 0.031310345977544785, 0.007866637781262398, 0.054129812866449356, -0.02245800010...
683
683
['Xiaoliang Dai', 'Hongxu Yin', 'Niraj K. Jha']
1711.02017v2
Neural networks (NNs) have begun to have a pervasive impact on various applications of machine learning. However, the problem of finding an optimal NN architecture for large applications has remained open for several decades. Conventional approaches search for the optimal NN architecture through extensive trial-and-err...
NeST: A Neural Network Synthesis Tool Based on a Grow-and-Prune Paradigm
2,017
http://arxiv.org/pdf/1711.02017v2
Title NeST Neural Network Synthesis Tool Based GrowandPrune Paradigm Summary Neural network NNs begun pervasive impact various application machine learning However problem finding optimal NN architecture large application remained open several decade Conventional approach search optimal NN architecture extensive triala...
[-0.04225454106926918, 0.045149169862270355, -0.055379170924425125, 0.027144653722643852, -0.03237073868513107, -0.04498456418514252, 0.04806896671652794, -0.011014611460268497, 0.014655706472694874, 0.025836342945694923, 0.0127420574426651, 0.005657525733113289, 0.005699627101421356, 0.08892931044101715, 0.02083749137...
684
684
['Filipe Rolim Cordeiro', 'Wellington Pinheiro dos Santos', 'Abel Guilhermino da Silva Filho']
1712.07312v1
Breast cancer is already one of the most common form of cancer worldwide. Mammography image analysis is still the most effective diagnostic method to promote the early detection of breast cancer. Accurately segmenting tumors in digital mammography images is important to improve diagnosis capabilities of health speciali...
Analysis of supervised and semi-supervised GrowCut applied to segmentation of masses in mammography images
2,017
http://arxiv.org/pdf/1712.07312v1
Title Analysis supervised semisupervised GrowCut applied segmentation mass mammography image Summary Breast cancer already one common form cancer worldwide Mammography image analysis still effective diagnostic method promote early detection breast cancer Accurately segmenting tumor digital mammography image important i...
[0.016599994152784348, -0.0029411392752081156, -0.023990413174033165, 0.001886810059659183, -0.05486414209008217, -0.0001775944692781195, 0.03415572643280029, 0.037300389260053635, 0.02353079617023468, 0.031302161514759064, 0.07191748917102814, 0.01975228823721409, -0.01802855171263218, 0.07661595195531845, -0.02419896...
685
685
['Shumeet Baluja']
1801.05156v1
We present extensive experiments training and testing hidden units in deep networks that emit only a predefined, static, number of discretized values. These units provide benefits in real-world deployment in systems in which memory and/or computation may be limited. Additionally, they are particularly well suited for u...
Empirical Explorations in Training Networks with Discrete Activations
2,018
http://arxiv.org/pdf/1801.05156v1
Title Empirical Explorations Training Networks Discrete Activations Summary present extensive experiment training testing hidden unit deep network emit predefined static number discretized value unit provide benefit realworld deployment system memory andor computation may limited Additionally particularly well suited u...
[-0.012845207937061787, 0.0042851511389017105, -0.014024388045072556, 0.0053544011898338795, 0.03046133928000927, -0.0171323474496603, 0.07774453610181808, -0.008793994784355164, -0.017757412046194077, 0.0006350188632495701, 0.010669970884919167, -0.01071455143392086, -0.007712224032729864, 0.038528744131326675, 0.0219...
686
686
['Esteban Real', 'Alok Aggarwal', 'Yanping Huang', 'Quoc V Le']
1802.01548v3
The effort devoted to hand-crafting image classifiers has motivated the use of architecture search to discover them automatically. Reinforcement learning and evolution have both shown promise for this purpose. This study employs a regularized version of a popular asynchronous evolutionary algorithm. We rigorously compa...
Regularized Evolution for Image Classifier Architecture Search
2,018
http://arxiv.org/pdf/1802.01548v3
Title Regularized Evolution Image Classifier Architecture Search Summary effort devoted handcrafting image classifier motivated use architecture search discover automatically Reinforcement learning evolution shown promise purpose study employ regularized version popular asynchronous evolutionary algorithm rigorously co...
[0.022280516102910042, 0.049255795776844025, -0.04031810164451599, 0.014892525039613247, -0.008276914246380329, -0.026720911264419556, 0.013798962347209454, 0.031990114599466324, -0.031773824244737625, 0.028270630165934563, -0.0018290071748197079, 0.04984038323163986, -0.015239505097270012, 0.04911781847476959, -0.0088...
687
687
['Alexander Wong', 'Mohammad Javad Shafiee', 'Francis Li', 'Brendan Chwyl']
1802.06488v1
Object detection is a major challenge in computer vision, involving both object classification and object localization within a scene. While deep neural networks have been shown in recent years to yield very powerful techniques for tackling the challenge of object detection, one of the biggest challenges with enabling ...
Tiny SSD: A Tiny Single-shot Detection Deep Convolutional Neural Network for Real-time Embedded Object Detection
2,018
http://arxiv.org/pdf/1802.06488v1
Title Tiny SSD Tiny Singleshot Detection Deep Convolutional Neural Network Realtime Embedded Object Detection Summary Object detection major challenge computer vision involving object classification object localization within scene deep neural network shown recent year yield powerful technique tackling challenge object...
[-0.00894768163561821, -0.015540339052677155, -0.010371421463787556, 0.09251029044389725, -0.0014948333846405149, -0.029780378565192223, 0.06379145383834839, -0.010930354706943035, -0.016195353120565414, -0.026631584390997887, 0.03221234679222107, 0.03448629751801491, -0.0014456340577453375, 0.029058344662189484, 0.014...
688
688
['Tobias Hinz', 'Stefan Wermter']
1803.02627v1
Combining Generative Adversarial Networks (GANs) with encoders that learn to encode data points has shown promising results in learning data representations in an unsupervised way. We propose a framework that combines an encoder and a generator to learn disentangled representations which encode meaningful information a...
Inferencing Based on Unsupervised Learning of Disentangled Representations
2,018
http://arxiv.org/pdf/1803.02627v1
Title Inferencing Based Unsupervised Learning Disentangled Representations Summary Combining Generative Adversarial Networks GANs encoders learn encode data point shown promising result learning data representation unsupervised way propose framework combine encoder generator learn disentangled representation encode mea...
[-0.03528452292084694, 0.12660232186317444, -0.018537303432822227, 0.07999861240386963, 0.01150921918451786, 0.03174477815628052, -0.004988578148186207, 0.017566945403814316, -0.007421194110065699, 0.03771132230758667, -0.01803317666053772, 0.05032367631793022, -0.010365442372858524, 0.03543653339147568, 0.061726044863...
689
689
['Erik Berglund', 'Joaquin Sitte']
0705.0199v2
The Parameter-Less Self-Organizing Map (PLSOM) is a new neural network algorithm based on the Self-Organizing Map (SOM). It eliminates the need for a learning rate and annealing schemes for learning rate and neighbourhood size. We discuss the relative performance of the PLSOM and the SOM and demonstrate some tasks in w...
The Parameter-Less Self-Organizing Map algorithm
2,007
http://arxiv.org/pdf/0705.0199v2
Title ParameterLess SelfOrganizing Map algorithm Summary ParameterLess SelfOrganizing Map PLSOM new neural network algorithm based SelfOrganizing Map SOM eliminates need learning rate annealing scheme learning rate neighbourhood size discus relative performance PLSOM SOM demonstrate task SOM fails PLSOM performs satisf...
[-0.03741709142923355, -0.0019337277626618743, 0.001419220701791346, -0.015870966017246246, 0.009401226416230202, -0.022940117865800858, 0.03638525679707527, -0.050921302288770676, 0.029635833576321602, 0.007876950316131115, -0.0322907380759716, 0.07134104520082474, 0.0012333877384662628, 0.015537651255726814, 0.017399...
690
690
['Christian Napoli', 'Giuseppe Pappalardo', 'Emiliano Tramontana', 'Zbigniew Marszałek', 'Dawid Połap', 'Marcin Woźniak']
1412.6464v1
In order to identify an object, human eyes firstly search the field of view for points or areas which have particular properties. These properties are used to recognise an image or an object. Then this process could be taken as a model to develop computer algorithms for images identification. This paper proposes the id...
Simplified firefly algorithm for 2D image key-points search
2,014
http://arxiv.org/pdf/1412.6464v1
Title Simplified firefly algorithm 2D image keypoints search Summary order identify object human eye firstly search field view point area particular property property used recognise image object process could taken model develop computer algorithm image identification paper proposes idea applying simplified firefly alg...
[-0.00666049076244235, -0.007869566790759563, -0.003758190432563424, 0.035692159086465836, -0.020446600392460823, -0.0026361218187958, 3.1921383197186515e-05, 0.037386175245046616, 0.006876166444271803, -0.03355340287089348, 0.014403634704649448, 0.05032142996788025, 0.02732980065047741, -0.0015734510961920023, 0.00337...
691
691
['Sue Han Lee', 'Chee Seng Chan', 'Paul Wilkin', 'Paolo Remagnino']
1506.08425v1
This paper studies convolutional neural networks (CNN) to learn unsupervised feature representations for 44 different plant species, collected at the Royal Botanic Gardens, Kew, England. To gain intuition on the chosen features from the CNN model (opposed to a 'black box' solution), a visualisation technique based on t...
Deep-Plant: Plant Identification with convolutional neural networks
2,015
http://arxiv.org/pdf/1506.08425v1
Title DeepPlant Plant Identification convolutional neural network Summary paper study convolutional neural network CNN learn unsupervised feature representation 44 different plant specie collected Royal Botanic Gardens Kew England gain intuition chosen feature CNN model opposed black box solution visualisation techniqu...
[0.006776423193514347, 0.017540596425533295, -0.013493082486093044, 0.09937453269958496, -0.015553660690784454, -0.001622108044102788, 0.02896924503147602, 0.015895109623670578, -0.029217569157481194, 0.021072691306471825, 0.00834373477846384, 0.033706702291965485, -0.00636943057179451, 0.03164862096309662, -0.00415315...
692
692
['Joshua C. Peterson', 'Joshua T. Abbott', 'Thomas L. Griffiths']
1608.02164v1
Deep neural networks have become increasingly successful at solving classic perception problems such as object recognition, semantic segmentation, and scene understanding, often reaching or surpassing human-level accuracy. This success is due in part to the ability of DNNs to learn useful representations of high-dimens...
Adapting Deep Network Features to Capture Psychological Representations
2,016
http://arxiv.org/pdf/1608.02164v1
Title Adapting Deep Network Features Capture Psychological Representations Summary Deep neural network become increasingly successful solving classic perception problem object recognition semantic segmentation scene understanding often reaching surpassing humanlevel accuracy success due part ability DNNs learn useful r...
[0.013091064058244228, 0.07363045960664749, -0.03308088704943657, 0.0372568815946579, -0.01791868545114994, 0.02727382630109787, 0.03783806413412094, 0.00030163361225277185, -0.021288791671395302, -0.01098821684718132, -0.06715395301580429, -0.002083749743178487, 0.015625564381480217, 0.05016550421714783, 0.01228214800...
693
693
['Esteban Real', 'Sherry Moore', 'Andrew Selle', 'Saurabh Saxena', 'Yutaka Leon Suematsu', 'Jie Tan', 'Quoc Le', 'Alex Kurakin']
1703.01041v2
Neural networks have proven effective at solving difficult problems but designing their architectures can be challenging, even for image classification problems alone. Our goal is to minimize human participation, so we employ evolutionary algorithms to discover such networks automatically. Despite significant computati...
Large-Scale Evolution of Image Classifiers
2,017
http://arxiv.org/pdf/1703.01041v2
Title LargeScale Evolution Image Classifiers Summary Neural network proven effective solving difficult problem designing architecture challenging even image classification problem alone goal minimize human participation employ evolutionary algorithm discover network automatically Despite significant computational requi...
[-0.011365228332579136, 0.07445085048675537, -0.04380756616592407, 0.036304619163274765, 0.010352357290685177, 0.0017203945899382234, 0.03758206218481064, 0.03093908540904522, -0.012030882760882378, 0.019904321059584618, 0.005386252887547016, 0.03029400296509266, -0.007501552812755108, 0.0726308524608612, 0.02211262099...
694
694
['Chunpeng Wu', 'Wei Wen', 'Tariq Afzal', 'Yongmei Zhang', 'Yiran Chen', 'Hai Li']
1703.04071v4
Recently, DNN model compression based on network architecture design, e.g., SqueezeNet, attracted a lot attention. No accuracy drop on image classification is observed on these extremely compact networks, compared to well-known models. An emerging question, however, is whether these model compression techniques hurt DN...
A Compact DNN: Approaching GoogLeNet-Level Accuracy of Classification and Domain Adaptation
2,017
http://arxiv.org/pdf/1703.04071v4
Title Compact DNN Approaching GoogLeNetLevel Accuracy Classification Domain Adaptation Summary Recently DNN model compression based network architecture design eg SqueezeNet attracted lot attention accuracy drop image classification observed extremely compact network compared wellknown model emerging question however w...
[-0.012206400744616985, 0.05085603520274162, -0.028635334223508835, 0.0407564640045166, -0.010684553533792496, 0.003392936196178198, 0.05592728778719902, 0.02588380128145218, -0.03154008463025093, 0.02351333387196064, -0.022899260744452477, 0.01830415055155754, 0.018779441714286804, 0.07889341562986374, 0.0061516696587...
695
695
['Mandar Haldekar', 'Ashwinkumar Ganesan', 'Tim Oates']
1706.04215v1
Traditional approaches to building a large scale knowledge graph have usually relied on extracting information (entities, their properties, and relations between them) from unstructured text (e.g. Dbpedia). Recent advances in Convolutional Neural Networks (CNN) allow us to shift our focus to learning entities and relat...
Identifying Spatial Relations in Images using Convolutional Neural Networks
2,017
http://arxiv.org/pdf/1706.04215v1
Title Identifying Spatial Relations Images using Convolutional Neural Networks Summary Traditional approach building large scale knowledge graph usually relied extracting information entity property relation unstructured text eg Dbpedia Recent advance Convolutional Neural Networks CNN allow u shift focus learning entit...
[0.04092583432793617, 0.046911127865314484, -0.0037572721485048532, 0.0848197489976883, -0.026717787608504295, 0.005975028499960899, 0.010528143495321274, 0.008887460455298424, 0.015241685323417187, -0.03622554987668991, -0.016750892624258995, 0.05479459837079048, -0.003973840735852718, 0.024060949683189392, 0.03149246...
696
696
['Adam R. Kosiorek', 'Alex Bewley', 'Ingmar Posner']
1706.09262v2
Class-agnostic object tracking is particularly difficult in cluttered environments as target specific discriminative models cannot be learned a priori. Inspired by how the human visual cortex employs spatial attention and separate "where" and "what" processing pathways to actively suppress irrelevant visual features, t...
Hierarchical Attentive Recurrent Tracking
2,017
http://arxiv.org/pdf/1706.09262v2
Title Hierarchical Attentive Recurrent Tracking Summary Classagnostic object tracking particularly difficult cluttered environment target specific discriminative model cannot learned priori Inspired human visual cortex employ spatial attention separate processing pathway actively suppress irrelevant visual feature work...
[0.01399760227650404, 0.014225639402866364, 0.017151782289147377, 0.07335303723812103, 0.02937290258705616, -0.009481535293161869, -0.004146979656070471, -0.015994062647223473, -0.03165788576006889, -0.03880343213677406, 0.004369891248643398, -0.02563496306538582, -0.006857514381408691, 0.0307307206094265, 0.0323091819...
697
697
['Seungkyun Hong', 'Seongchan Kim', 'Minsu Joh', 'Sa-kwang Song']
1711.10644v2
Predicting unseen weather phenomena is an important issue for disaster management. In this paper, we suggest a model for a convolutional sequence-to-sequence autoencoder for predicting undiscovered weather situations from previous satellite images. We also propose a symmetric skip connection between encoder and decoder...
PSIque: Next Sequence Prediction of Satellite Images using a Convolutional Sequence-to-Sequence Network
2,017
http://arxiv.org/pdf/1711.10644v2
Title PSIque Next Sequence Prediction Satellite Images using Convolutional SequencetoSequence Network Summary Predicting unseen weather phenomenon important issue disaster management paper suggest model convolutional sequencetosequence autoencoder predicting undiscovered weather situation previous satellite image also ...
[0.018754642456769943, 0.09524159133434296, -0.001475048717111349, 0.04831339046359062, 0.0031933083664625883, 0.010652001947164536, -0.0016888133250176907, -0.025227637961506844, -0.002432959619909525, 0.02826281450688839, 0.0632391944527626, -0.02445962280035019, 0.020399456843733788, 0.09698973596096039, 0.029526859...
698
698
['Wellington Pinheiro dos Santos', 'Ricardo Emmanuel de Souza', 'Plínio B. dos Santos Filho']
1712.00712v1
Alzheimer's disease is the most common cause of dementia, yet hard to diagnose precisely without invasive techniques, particularly at the onset of the disease. This work approaches image analysis and classification of synthetic multispectral images composed by diffusion-weighted magnetic resonance (MR) cerebral images ...
Evaluation of Alzheimer's Disease by Analysis of MR Images using Multilayer Perceptrons and Kohonen SOM Classifiers as an Alternative to the ADC Maps
2,017
http://arxiv.org/pdf/1712.00712v1
Title Evaluation Alzheimers Disease Analysis MR Images using Multilayer Perceptrons Kohonen SOM Classifiers Alternative ADC Maps Summary Alzheimers disease common cause dementia yet hard diagnose precisely without invasive technique particularly onset disease work approach image analysis classification synthetic multis...
[-0.04594212397933006, 0.01745085045695305, -0.030492210760712624, 0.013584626838564873, 0.01706668548285961, 0.012701579369604588, 0.02677692100405693, 0.022421682253479958, 0.09287978708744049, 0.026735689491033554, 0.018283044919371605, 0.016190290451049805, 0.02528703771531582, 0.01034840289503336, 0.03909936919808...
699
699
['Cheston Tan', 'Tomaso Poggio']
1406.3793v1
Faces are a class of visual stimuli with unique significance, for a variety of reasons. They are ubiquitous throughout the course of a person's life, and face recognition is crucial for daily social interaction. Faces are also unlike any other stimulus class in terms of certain physical stimulus characteristics. Furthe...
Neural tuning size is a key factor underlying holistic face processing
2,014
http://arxiv.org/pdf/1406.3793v1
Title Neural tuning size key factor underlying holistic face processing Summary Faces class visual stimulus unique significance variety reason ubiquitous throughout course person life face recognition crucial daily social interaction Faces also unlike stimulus class term certain physical stimulus characteristic Further...
[0.016153737902641296, -0.029728543013334274, -0.04741014167666435, 0.016164788976311684, 0.02687034383416176, 0.053367383778095245, 0.05962429940700531, 0.01388290524482727, -0.018673652783036232, 0.012384072877466679, -0.03280370309948921, -0.0035295968409627676, 0.045720186084508896, 0.08104520291090012, 0.082537136...