Lecture Notes in Computer Science 12396

Founding Editors

Gerhard GoosKarlsruhe Institute of Technology, Karlsruhe, Germany

Juris HartmanisCornell University, Ithaca, NY, USA

Editorial Board Members

Elisa BertinoPurdue University, West Lafayette, IN, USA

Wen GaoPeking University, Beijing, China

Bernhard SteffenTU Dortmund University, Dortmund, Germany

Gerhard WoegingerRWTH Aachen, Aachen, Germany

Moti YungColumbia University, New York, NY, USA

More information about this series at http://www.springer.com/series/7407

Igor Farkaš • Paolo Masulli •

Stefan Wermter (Eds.)

Artificial Neural Networksand Machine Learning –

ICANN 202029th International Conference on Artificial Neural NetworksBratislava, Slovakia, September 15–18, 2020Proceedings, Part I

123

EditorsIgor FarkašDepartment of Applied InformaticsComenius University in BratislavaBratislava, Slovakia

Paolo MasulliDepartment of Applied Mathematicsand Computer ScienceTechnical University of DenmarkKgs. Lyngby, Denmark

Stefan WermterDepartment of InformaticsUniversity of HamburgHamburg, Germany

ISSN 0302-9743 ISSN 1611-3349 (electronic)Lecture Notes in Computer ScienceISBN 978-3-030-61608-3 ISBN 978-3-030-61609-0 (eBook)https://doi.org/10.1007/978-3-030-61609-0

LNCS Sublibrary: SL1 – Theoretical Computer Science and General Issues

© Springer Nature Switzerland AG 2020This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of thematerial is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,broadcasting, reproduction on microfilms or in any other physical way, and transmission or informationstorage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology nowknown or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoes not imply, even in the absence of a specific statement, that such names are exempt from the relevantprotective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in this book arebelieved to be true and accurate at the date of publication. Neither the publisher nor the authors or the editorsgive a warranty, expressed or implied, with respect to the material contained herein or for any errors oromissions that may have been made. The publisher remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AGThe registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

Research on artificial neural networks has progressed over decades, in recent yearsbeing fueled especially by deep learning that has proven to be data-greedy but efficientin solving various, mostly supervised tasks. Applications of artificial neural networks,especially related to artificial intelligence, influence our lives, reaching new horizons.Examples range from autonomous car driving, virtual assistants, and decision supportsystems, to healthcare data analytics, financial forecasting, and smart devices in ourhomes, just to name a few. These developments, however, also provide challenges,which were not imaginable previously, e.g., verification to prevent manipulation ofvoice, videos, or people’s opinions during elections.

The International Conference on Artificial Neural Networks (ICANN) is the annualflagship conference of the European Neural Network Society (ENNS). This year, thespecial situation due to the COVID-19 pandemic influenced the planning of thisconference in an unprecedented way. Based on the restrictions to travel and gatherings,as well as forecasts for the following months, it was not appropriate to decide for alarge gathering such as ICANN to take place as a physical event in September 2020.Therefore, after a lot of consideration and discussions, the Organizing Committee,together with the Executive Committee of ENNS decided to postpone the physicalmeeting of ICANN 2020 and schedule next year’s ICANN in September 2021 inBratislava, Slovakia, since we also believe that a physical meeting has so manyadvantages compared to a virtual one.

Nevertheless, we decided to assist and do justice to the authors by disseminatingtheir current work already completed for this year, running the paper selection andreview process so that the successful submissions could appear online. Following along-standing successful collaboration, the proceedings of ICANN are published asvolumes within Springer’s Lecture Notes in Computer Science series. The response tothis year’s call for papers resulted in an impressive number of 381 article submissions,of which almost all were long papers. After the official announcement that the con-ference would be postponed, 21 authors decided to withdraw their papers. The paperselection and review process that followed was decided during the online meetingof the Bratislava organizing team and the ENNS Executive Committee. The 20 Pro-gram Committee (PC) members agreed to review the long papers in stages. Stage oneinvolved the review of 249 papers, of which 188 papers were selected for a stage-tworeview by independent reviewers. The majority of PC members had doctoral degrees(85%) and 70% of them were also professors. At the stage-two review, in total, 154reviewers participated in the process, all having filled in an online questionnairefocusing on their areas of expertise, which significantly helped the general chair toproperly assign papers to them. The reviewers were assigned one to three articles, buteach article received three reports by the PC and reviewers, and these served as a majorsource for the final decision.

In total, 142 articles were accepted for the proceedings and the authors wererequested to submit final versions. The acceptance rate was hence 37% when calculatedfrom all initial submissions, or 57% when calculated from the initial papers selected forstage-two reviews. A list of PC members and reviewers, who agreed to publish theirnames, is included in these proceedings. With these procedures we tried to keep thequality of the proceedings high, while still having a critical mass of contributionsreflecting the progress of the field. Overall, we hope that these proceedings will con-tribute to the dissemination of new results by the neural network community duringthese challenging times and that we can again have a physical ICANN in 2021.

We greatly appreciate the PC members and the reviewers for their invaluable work.

September 2020 Igor FarkašPaolo Masulli

Stefan Wermter

vi Preface

Organization

General Chairs

Igor Farkaš Comenius University in Bratislava, SlovakiaL’ubica Beñušková Comenius University in Bratislava, Slovakia

Organizing Committee Chairs

Kristína Malinovská Comenius University in Bratislava, SlovakiaAlessandra Lintas ENNS Lausanne, Switzerland

Honorary Chairs

Stefan Wermter University of Hamburg, GermanyVěra Kůrková Czech Academy of Sciences, Czech Republic

Program Committee

L’ubica Beňušková Comenius University in Bratislava, SlovakiaJérémie Cabessa Panthéon-Assas University Paris II, FranceWlodek Duch Nicolaus Copernicus University, PolandIgor Farkaš Comenius University in Bratislava, SlovakiaJuraj Holas Comenius University in Bratislava, SlovakiaVěra Kůrková Czech Academy of Sciences, Czech RepublicTomáš Kuzma Comenius University in Bratislava, SlovakiaAlessandra Lintas University of Lausanne, SwitzerlandKristína Malinovská Comenius University in Bratislava, SlovakiaPaolo Masulli Technical University of Denmark, DenmarkAlessio Micheli University of Pisa, ItalySebastian Otte University of Tübingen, GermanyJaakko Peltonen University of Tampere, FinlandAntonio J. Pons University of Barcelona, SpainMartin Takáč Comenius University in Bratislava, SlovakiaIgor V. Tetko Technical University Munich, GermanyMatúš Tuna Comenius University in Bratislava, SlovakiaAlessandro E. P. Villa University of Lausanne, SwitzerlandRoseli Wedemann Rio de Janeiro State University, BrazilStefan Wermter University of Hamburg, Germany

Communication Chair

Paolo Masulli ENNS, Technical University of Denmark, Denmark

Reviewers

Argimiro Arratia Polytechnic University of Catalonia, SpainAndrá Artelt Bielefeld University, GermanyMiguel Atencia Universidad de Malaga, SpainCristian Axenie Huawei German Research Center, GermanyFatemeh Azimi TU Kaiserslautern, GermanyJatin Bedi BITS Pilani, IndiaL’ubica Beňušková Comenius University in Bratislava, SlovakiaBernhard Bermeitinger Universität St. Gallen, SwitzerlandYann Bernard Inria, FranceJyostna Devi Bodapati Indian Institute of Technology Madras, IndiaNicolas Bougie National Institute of Informatics, JapanEvgeny Burnaev Skoltech, RussiaRüdiger Busche Osnabrück University, GermanyJérémie Cabessa Panthéon-Assas University Paris II, FranceHugo Eduardo Camacho Universidad Autónoma de Tamaulipas, MexicoAntonio Candelieri University of Milano-Bicocca, ItalySiyu Cao Beijing Jiaotong University, ChinaAntonio Carta University of Pisa, ItalyNico Cavalcanti UFPE, BrazilGavneet Singh Chadha South Westphalia University of Applied Sciences,

GermanyShengjia Chen Guangxi Normal University, ChinaAlessandro Di Nuovo Sheffield Hallam University, UKTayssir Doghri INRS, CanadaHaizhou Du Shanghai University of Electric Power, ChinaWlodzislaw Duch Nicolaus Copernicus University, PolandOla Engkvist AstraZeneca, SwedenManfred Eppe University of Hamburg, GermanyYuchun Fang Shanghai University, ChinaIgor Farkaš Comenius University in Bratislava, SlovakiaOliver Gallitz Technische Hochschule Ingolstadt, GermanyJochen Garcke University of Bonn, GermanyDominik Geissler Relayr GmbH, GermanyClaudio Giorgio

GiancaterinoCatholic University of Milan, Italy

Francesco Giannini University of Siena, ItalyKathrin Grosse CISPA Helmholtz Center for Information Security,

GermanyPhilipp Grüning University of Lübeck, GermanyMichael Guckert Technische Hochschule Mittelhessen, Germany

viii Organization

Alberto Guillén University of Granada, SpainSong Guo Nankai University, ChinaSimon Hakenes Ruhr-Universität Bochum, GermanyXiaoxu Han Tianjin University, ChinaMartina Hasenjäger Honda Research Institute Europe GmbH, GermanyTieke He Nanjing University, ChinaRaoul Heese Fraunhofer ITWM, GermanyXavier Hinaut Inria, FranceJuraj Holas Comenius University in Bratislava, SlovakiaJunjie Huang Chinese Academy of Sciences, ChinaDania Humaidan University of Tübingen, GermanyNicolangelo Iannella University of Oslo, NorwayNoman Javed London School of Economics, UKShaoxiong Ji Aalto University, FinlandDoreen Jirak University of Hamburg, GermanyRenaud Jolivet University of Geneva, SwitzerlandJan Kalina Czech Academy of Sciences, Czech RepublicIzumi Karino The University of Tokyo, JapanJohn Kelleher Technological University Dublin, IrelandMatthias Kerzel University of Hamburg, GermanyAdil Khan Innopolis University, RussiaMatthias Kissel Technical University of Munich, GermanyAtsushi Koike National Institute of Technology, JapanStefanos Kollias NTUA, GreeceEkaterina Komendantskaya Heriot–Watt University, UKPetia Koprinkova-Hristova IICT–BAS, BulgariaIrena Koprinska The University of Sydney, AustraliaConstantine Kotropoulos Aristotle University of Thessaloniki, GreeceAdam Krzyzak Concordia University, CanadaVěra Kůrková Czech Academy of Sciences, Czech RepublicSumit Kushwaha Kamla Nehru Institute of Technology, IndiaTomáš Kuzma Comenius University in Bratislava, SlovakiaNicolas Lachiche University of Strasbourg, FranceWai Lam The Chinese University of Hong Kong, Hong KongYajie Li Chinese Academy of Sciences, ChinaMingxing Li Guangxi Normal University, ChinaMengdi Li University of Hamburg, GermanyBoquan Li Chinese Academy of Sciences, ChinaJianfeng Li Southwest University, ChinaYang Lin The University of Sydney, AustraliaAlessandra Lintas University of Lausanne, SwitzerlandYezheng Liu Hefei University of Techonology, ChinaVadim Liventsev TU Eindhoven, The NetherlandsViktor Liviniuk University of California Irvine, USANasrulloh Loka Ghent University, BelgiumShuai Lu Jilin University, China

Organization ix

An Luo South China University of Technology, ChinaThomas Lymburn The University of Western Australia, AustraliaKleanthis Malialis University of Cyprus, CyprusKristína Malinovská Comenius University in Bratislava, SlovakiaFragkiskos Malliaros CentraleSupélec, FranceGilles Marcou University of Strasbourg, FranceMichael Marino Universität Osnabrück, GermanyPaolo Masulli Technical University of Denmark, DenmarkGuillaume Matheron Institut des Systèmes Intelligents et de Robotique,

FranceAlessio Micheli University of Pisa, ItalyFlorian Mirus BMW Group, GermanyRoman Neruda Institute of Computer Science, ASCR, Czech RepublicHasna Njah University of Sfax, TunisiaMihaela Oprea Petroleum-Gas University of Ploiesti, RomaniaChristoph Ostrau Bielefeld University, GermanySebastian Otte University of Tübingen, GermanyHyeyoung Park Kyungpook National University Daegu Campus,

South KoreaJaakko Peltonen Tampere University, FinlandDaniele Perlo University of Turin, ItalyVincenzo Piuri University of Milan, ItalyAntonio Javier Pons Rivero Universitat Politècnica de Catalunya, SpainMike Preuss Universiteit Leiden, The NetherlandsMiloš Prágr Czech Technical University, FEE, Czech RepublicYili Qu Sun Yat-Sen University, ChinaLaya Rafiee Concordia University, CanadaRajkumar Ramamurthy Fraunhofer IAIS, GermanyZuzana Roštáková Slovak Academy of Sciences, Slovakia

Frank Röder University of Hamburg, GermanyJun Sang Chongqing University, ChinaAnindya Sarkar Indian Institute of Technology, IndiaYikemaiti Sataer Southeast University, ChinaSimone Scardapane Sapienza University of Rome, ItalyJochen Schmidt Rosenheim University of Applied Sciences, GermanyCedric Schockaert Paul Wurth Geprolux S.A., LuxembourgFriedhelm Schwenker University of Ulm, GermanyAndreas Sedlmeier LMU Munich, GermanyGabriela Šejnová Czech Technical University in Prague, Czech RepublicAlexandru Serban Radboud University, The NetherlandsLinlin Shen Shenzhen University, ChinaShashwat Shukla Indian Institute of Technology Bombay, IndiaCaio Silva Universidade Federal de Pernambuco, BrazilAleksander Smywiński-Pohl AGH University of Science and Technology, PolandPouya Soltani Zarrin IHP, Germany

x Organization

Miguel Soriano Institute for Cross-Disciplinary Physics and ComplexSystems, Spain

Lea Steffen FZI Research Center for Information Technology,Germany

Michael Stettler University of Tübingen, GermanyRuxandra Stoean University of Craiova, RomaniaJérémie Sublime ISEP, FranceChanchal Suman IIT Patna, IndiaXiaoqi Sun Shanghai University, ChinaAlexander Sutherland University of Hamburg, GermanyZaneta Swiderska-Chadaj Warsaw University of Technology, PolandRudolf Szadkowski Czech Technical University in Prague, Czech RepublicPhilippe Thomas Université de Lorraine, FranceShiro Takagi The University of Tokyo, JapanMartin Takáč Comenius University, SlovakiaMax Talanov Kazan Federal University, RussiaEnzo Tartaglione Università degli Studi Torino, ItalyIgor Tetko Helmholtz Zentrum München, GermanyJuan-Manuel

Torres-MorenoUniversité d’Avignon, France

Jochen Triesch Frankfurt Institute for Advanced Studies, GermanyMatúš Tuna Comenius University in Bratislava, SlovakiaTakaya Ueda Ritsumeikan University, JapanSagar Verma CentraleSupélec, FranceRicardo Vigário Universidade NOVA de Lisboa, PortugalAlessandro E. P. Villa University of Lausanne, SwitzerlandPaolo Viviani Noesis Solutions NV, BelgiumShuo Wang Monash University and CSIRO, AustraliaHuiling Wang Tampere University, FinlandXing Wang Ningxia University, ChinaZhe Wang Soochow University, ChinaRoseli Wedemann Rio de Janeiro State University, BrazilBaole Wei Chinese Academy of Sciences, ChinaFeng Wei York University, CanadaYingcan Wei The University of Hong Kong, Hong KongMartin Georg Weiß Regensburg University of Applied Sciences, GermanyThomas Wennekers Plymouth University, UKMarc Wenninger Rosenheim Technical University of Applied Sciences,

GermanyStefan Wermter University of Hamburg, GermanyJohn Wilmes Brandeis University, USAChristoph Windheuser ThoughtWorks Inc., GermanyMoritz Wolter University of Bonn, GermanyChangmin Wu Ecole Polytechnique, FranceTakaharu Yaguchi Kobe University, JapanTsoy Yury Solidware, South Korea

Organization xi

Yuchen Zheng Kyushu University, JapanMeng Zhou Shanghai Jiao Tong University, ChinaCongcong Zhu Shanghai University, ChinaAndrea Zugarini University of Florence, ItalyAdrian De Wynter Amazon Alexa, USAZhao Jiapeng Chinese Academy of Sciences, ChinaTessa Van Der Heiden BMW Group, GermanyTadeusz Wieczorek Silesian University of Technology, Poland

xii Organization

Contents – Part I

Adversarial Machine Learning

On the Security Relevance of Initial Weights in Deep Neural Networks. . . . . 3Kathrin Grosse, Thomas A. Trost, Marius Mosbach, Michael Backes,and Dietrich Klakow

Fractal Residual Network for Face Image Super-Resolution . . . . . . . . . . . . . 15Yuchun Fang, Qicai Ran, and Yifan Li

From Imbalanced Classification to Supervised Outlier Detection Problems:Adversarially Trained Auto Encoders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Max Lübbering, Rajkumar Ramamurthy, Michael Gebauer, Thiago Bell,Rafet Sifa, and Christian Bauckhage

Generating Adversarial Texts for Recurrent Neural Networks . . . . . . . . . . . . 39Chang Liu, Wang Lin, and Zhengfeng Yang

Enforcing Linearity in DNN Succours Robustness and AdversarialImage Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Anindya Sarkar and Raghu Iyengar

Computational Analysis of Robustness in Neural Network Classifiers . . . . . . 65Iveta Bečková, Štefan Pócoš, and Igor Farkaš

Bioinformatics and Biosignal Analysis

Convolutional Neural Networks with Reusable Full-Dimension-LongLayers for Feature Selection and Classification of Motor Imageryin EEG Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Mikhail Tokovarov

Compressing Genomic Sequences by Using Deep Learning . . . . . . . . . . . . . 92Wenwen Cui, Zhaoyang Yu, Zhuangzhuang Liu, Gang Wang,and Xiaoguang Liu

Learning Tn5 Sequence Bias from ATAC-seq on Naked Chromatin . . . . . . . 105Meshal Ansari, David S. Fischer, and Fabian J. Theis

Tucker Tensor Decomposition of Multi-session EEG Data . . . . . . . . . . . . . . 115Zuzana Rošťáková, Roman Rosipal, and Saman Seifpour

Reactive Hand Movements from Arm Kinematics and EMG Signals Basedon Hierarchical Gaussian Process Dynamical Models. . . . . . . . . . . . . . . . . . 127

Nick Taubert, Jesse St. Amand, Prerana Kumar, Leonardo Gizzi,and Martin A. Giese

Cognitive Models

Investigating Efficient Learning and Compositionality in GenerativeLSTM Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Sarah Fabi, Sebastian Otte, Jonas Gregor Wiese, and Martin V. Butz

Fostering Event Compression Using Gated Surprise . . . . . . . . . . . . . . . . . . 155Dania Humaidan, Sebastian Otte, and Martin V. Butz

Physiologically-Inspired Neural Circuits for the Recognitionof Dynamic Faces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

Michael Stettler, Nick Taubert, Tahereh Azizpour, Ramona Siebert,Silvia Spadacenta, Peter Dicke, Peter Thier, and Martin A. Giese

Hierarchical Modeling with Neurodynamical Agglomerative Analysis . . . . . . 180Michael Marino, Georg Schröter, Gunther Heidemann,and Joachim Hertzberg

Convolutional Neural Networks and Kernel Methods

Deep and Wide Neural Networks Covariance Estimation . . . . . . . . . . . . . . . 195Argimiro Arratia, Alejandra Cabaña, and José Rafael León

Monotone Deep Spectrum Kernels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207Ivano Lauriola and Fabio Aiolli

Permutation Learning in Convolutional Neural Networksfor Time-Series Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

Gavneet Singh Chadha, Jinwoo Kim, Andreas Schwung,and Steven X. Ding

Deep Learning Applications I

GTFNet: Ground Truth Fitting Network for Crowd Counting . . . . . . . . . . . . 235Jinghan Tan, Jun Sang, Zhili Xiang, Ying Shi, and Xiaofeng Xia

Evaluation of Deep Learning Methods for Bone Suppressionfrom Dual Energy Chest Radiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

Ilyas Sirazitdinov, Konstantin Kubrak, Semen Kiselev, Alexey Tolkachev,Maksym Kholiavchenko, and Bulat Ibragimov

xiv Contents – Part I

Multi-person Absolute 3D Human Pose Estimation with WeakDepth Supervision. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

Márton Véges and András Lőrincz

Solar Power Forecasting Based on Pattern Sequence Similarityand Meta-learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

Yang Lin, Irena Koprinska, Mashud Rana, and Alicia Troncoso

Analysis and Prediction of Deforming 3D Shapes Using Oriented BoundingBoxes and LSTM Autoencoders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284

Sara Hahner, Rodrigo Iza-Teran, and Jochen Garcke

Deep Learning Applications II

Novel Sketch-Based 3D Model Retrieval via Cross-domain FeatureClustering and Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

Kai Gao, Jian Zhang, Chen Li, Changbo Wang, Gaoqi He,and Hong Qin

Multi-objective Cuckoo Algorithm for Mobile Devices NetworkArchitecture Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312

Nan Zhang, Jianzong Wang, Jian Yang, Xiaoyang Qu, and Jing Xiao

DeepED: A Deep Learning Framework for EstimatingEvolutionary Distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

Zhuangzhuang Liu, Mingming Ren, Zhiheng Niu, Gang Wang,and Xiaoguang Liu

Interpretable Machine Learning Structure for an Early Predictionof Lane Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

Oliver Gallitz, Oliver De Candido, Michael Botsch, Ron Melz,and Wolfgang Utschick

Explainable Methods

Convex Density Constraints for Computing PlausibleCounterfactual Explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353

André Artelt and Barbara Hammer

Identifying Critical States by the Action-Based Varianceof Expected Return . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366

Izumi Karino, Yoshiyuki Ohmura, and Yasuo Kuniyoshi

Explaining Concept Drift by Mean of Direction . . . . . . . . . . . . . . . . . . . . . 379Fabian Hinder, Johannes Kummert, and Barbara Hammer

Contents – Part I xv

Few-Shot Learning

Context Adaptive Metric Model for Meta-learning . . . . . . . . . . . . . . . . . . . 393Zhe Wang and Fanzhang Li

Ensemble-Based Deep Metric Learning for Few-Shot Learning . . . . . . . . . . . 406Meng Zhou, Yaoyi Li, and Hongtao Lu

More Attentional Local Descriptors for Few-Shot Learning . . . . . . . . . . . . . 419Hui Li, Liu Yang, and Fei Gao

Implementation of Siamese-Based Few-Shot Learning Algorithmsfor the Distinction of COPD and Asthma Subjects . . . . . . . . . . . . . . . . . . . 431

Pouya Soltani Zarrin and Christian Wenger

Few-Shot Learning for Medical Image Classification . . . . . . . . . . . . . . . . . . 441Aihua Cai, Wenxin Hu, and Jun Zheng

Generative Adversarial Network

Adversarial Defense via Attention-Based Randomized Smoothing . . . . . . . . . 455Xiao Xu, Shiyu Feng, Zheng Wang, Lizhe Xie, and Yining Hu

Learning to Learn from Mistakes: Robust Optimizationfor Adversarial Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467

Alex Serban, Erik Poll, and Joost Visser

Unsupervised Anomaly Detection with a GAN Augmented Autoencoder . . . . 479Laya Rafiee and Thomas Fevens

An Efficient Blurring-Reconstruction Model to Defend AgainstAdversarial Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491

Wen Zhou, Liming Wang, and Yaohao Zheng

EdgeAugment: Data Augmentation by Fusing and Filling Edge Maps . . . . . . 504Bangfeng Xia, Yueling Zhang, Weiting Chen, Xiangfeng Wang,and Jiangtao Wang

Face Anti-spoofing with a Noise-Attention Network Using Color-ChannelDifference Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517

Yuanyuan Ren, Yongjian Hu, Beibei Liu, Yixiang Xie, and Yufei Wang

Generative and Graph Models

Variational Autoencoder with Global- and Medium Timescale Auxiliariesfor Emotion Recognition from Speech . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529

Hussam Almotlak, Cornelius Weber, Leyuan Qu, and Stefan Wermter

xvi Contents – Part I

Improved Classification Based on Deep Belief Networks . . . . . . . . . . . . . . . 541Jaehoon Koo and Diego Klabjan

Temporal Anomaly Detection by Deep Generative Modelswith Applications to Biological Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553

Takaya Ueda, Yukako Tohsato, and Ikuko Nishikawa

Inferring, Predicting, and Denoising Causal Wave Dynamics . . . . . . . . . . . . 566Matthias Karlbauer, Sebastian Otte, Hendrik P. A. Lensch,Thomas Scholten, Volker Wulfmeyer, and Martin V. Butz

PART-GAN: Privacy-Preserving Time-Series Sharing . . . . . . . . . . . . . . . . . 578Shuo Wang, Carsten Rudolph, Surya Nepal, Marthie Grobler,and Shangyu Chen

EvoNet: A Neural Network for Predicting the Evolutionof Dynamic Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594

Changmin Wu, Giannis Nikolentzos, and Michalis Vazirgiannis

Hybrid Neural-Symbolic Architectures

Facial Expression Recognition Method Based on a Part-Based TemporalConvolutional Network with a Graph-Structured Representation . . . . . . . . . . 609

Lei Zhong, Changmin Bai, Jianfeng Li, Tong Chen, and Shigang Li

Generating Facial Expressions Associated with Text . . . . . . . . . . . . . . . . . . 621Lisa Graziani, Stefano Melacci, and Marco Gori

Bilinear Fusion of Commonsense Knowledge with Attention-BasedNLI Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633

Amit Gajbhiye, Thomas Winterbottom, Noura Al Moubayed,and Steven Bradley

Neural-Symbolic Relational Reasoning on Graph Models: Effective LinkInference and Computation from Knowledge Bases . . . . . . . . . . . . . . . . . . . 647

Henrique Lemos, Pedro Avelar, Marcelo Prates, Artur Garcez,and Luís Lamb

Tell Me Why You Feel That Way: Processing Compositional Dependencyfor Tree-LSTM Aspect Sentiment Triplet Extraction (TASTE) . . . . . . . . . . . 660

Alexander Sutherland, Suna Bensch, Thomas Hellström, Sven Magg,and Stefan Wermter

SOM-Based System for Sequence Chunking and Planning . . . . . . . . . . . . . . 672Martin Takac, Alistair Knott, and Mark Sagar

Contents – Part I xvii

Image Processing

Bilinear Models for Machine Learning. . . . . . . . . . . . . . . . . . . . . . . . . . . . 687Tayssir Doghri, Leszek Szczecinski, Jacob Benesty, and Amar Mitiche

Enriched Feature Representation and Combination for DeepSaliency Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 699

Lecheng Zhou and Xiaodong Gu

Spectral Graph Reasoning Network for Hyperspectral Image Classification. . . 711Huiling Wang

Salient Object Detection with Edge Recalibration . . . . . . . . . . . . . . . . . . . . 724Zhenshan Tan, Yikai Hua, and Xiaodong Gu

Multi-Scale Cross-Modal Spatial Attention Fusion for Multi-labelImage Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 736

Junbing Li, Changqing Zhang, Xueman Wang, and Ling Du

A New Efficient Finger-Vein Verification Based on Lightweight NeuralNetwork Using Multiple Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748

Haocong Zheng, Yongjian Hu, Beibei Liu, Guang Chen, and Alex C. Kot

Medical Image Processing

SU-Net: An Efficient Encoder-Decoder Model of Federated Learningfor Brain Tumor Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 761

Liping Yi, Jinsong Zhang, Rui Zhang, Jiaqi Shi, Gang Wang,and Xiaoguang Liu

Synthesis of Registered Multimodal Medical Images with Lesions. . . . . . . . . 774Yili Qu, Wanqi Su, Xuan Lv, Chufu Deng, Ying Wang, Yutong Lu,Zhiguang Chen, and Nong Xiao

ACE-Net: Adaptive Context Extraction Network for MedicalImage Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787

Tuo Leng, Yu Wang, Ying Li, and Zhijie Wen

Wavelet U-Net for Medical Image Segmentation. . . . . . . . . . . . . . . . . . . . . 800Ying Li, Yu Wang, Tuo Leng, and Wen Zhijie

Recurrent Neural Networks

Character-Based LSTM-CRF with Semantic Features for ChineseEvent Element Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813

Wei Liu, Yusen Wu, Lei Jiang, Jianfeng Fu, and Weimin Li

xviii Contents – Part I

Sequence Prediction Using Spectral RNNs . . . . . . . . . . . . . . . . . . . . . . . . . 825Moritz Wolter, Jürgen Gall, and Angela Yao

Attention Based Mechanism for Load Time SeriesForecasting: AN-LSTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 838

Jatin Bedi

DartsReNet: Exploring New RNN Cells in ReNet Architectures . . . . . . . . . . 850Brian B. Moser, Federico Raue, Jörn Hees, and Andreas Dengel

On Multi-modal Fusion for Freehand Gesture Recognition . . . . . . . . . . . . . . 862Monika Schak and Alexander Gepperth

Recurrent Neural Network Learning of Performance and IntrinsicPopulation Dynamics from Sparse Neural Data . . . . . . . . . . . . . . . . . . . . . . 874

Alessandro Salatiello and Martin A. Giese

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887

Contents – Part I xix

Contents – Part II

Model Compression I

Fine-Grained Channel Pruning for Deep Residual Neural Networks. . . . . . . . 3Siang Chen, Kai Huang, Dongliang Xiong, Bowen Li, and Luc Claesen

A Lightweight Fully Convolutional Neural Network of High AccuracySurface Defect Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Yajie Li, Yiqiang Chen, Yang Gu, Jianquan Ouyang, Jiwei Wang,and Ni Zeng

Detecting Uncertain BNN Outputs on FPGA Using Monte CarloDropout Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Tomoyuki Myojin, Shintaro Hashimoto, and Naoki Ishihama

Neural Network Compression via Learnable Wavelet Transforms . . . . . . . . . 39Moritz Wolter, Shaohui Lin, and Angela Yao

Fast and Robust Compression of Deep Convolutional Neural Networks . . . . . 52Jia Wen, Liu Yang, and Chenyang Shen

Model Compression II

Pruning Artificial Neural Networks: A Way to Find Well-Generalizing,High-Entropy Sharp Minima. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Enzo Tartaglione, Andrea Bragagnolo, and Marco Grangetto

Log-Nets: Logarithmic Feature-Product Layers Yield MoreCompact Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Philipp Grüning, Thomas Martinetz, and Erhardt Barth

Tuning Deep Neural Network’s Hyperparameters Constrainedto Deployability on Tiny Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Riccardo Perego, Antonio Candelieri, Francesco Archetti,and Danilo Pau

Obstacles to Depth Compression of Neural Networks . . . . . . . . . . . . . . . . . 104Will Burstein and John Wilmes

Multi-task and Multi-label Learning

Multi-label Quadruplet Dictionary Learning . . . . . . . . . . . . . . . . . . . . . . . . 119Jiayu Zheng, Wencheng Zhu, and Pengfei Zhu

Pareto Multi-task Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132Salvatore D. Riccio, Deyan Dyankov, Giorgio Jansen,Giuseppe Di Fatta, and Giuseppe Nicosia

Convex Graph Laplacian Multi-Task Learning SVM . . . . . . . . . . . . . . . . . . 142Carlos Ruiz, Carlos M. Alaíz, and José R. Dorronsoro

Neural Network Theory and Information Theoretic Learning

Prediction Stability as a Criterion in Active Learning. . . . . . . . . . . . . . . . . . 157Junyu Liu, Xiang Li, Jiqiang Zhou, and Jianxiong Shen

Neural Spectrum Alignment: Empirical Study . . . . . . . . . . . . . . . . . . . . . . . 168Dmitry Kopitkov and Vadim Indelman

Nonlinear, Nonequilibrium Landscape Approach to NeuralNetwork Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

Roseli S. Wedemann and Angel R. Plastino

Hopfield Networks for Vector Quantization . . . . . . . . . . . . . . . . . . . . . . . . 192C. Bauckhage, R. Ramamurthy, and R. Sifa

Prototype-Based Online Learning on Homogeneously LabeledStreaming Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

Christian Limberg, Jan Philip Göpfert, Heiko Wersing,and Helge Ritter

Normalization and Regularization Methods

Neural Network Training with Safe Regularization in the Null Spaceof Batch Activations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

Matthias Kissel, Martin Gottwald, and Klaus Diepold

The Effect of Batch Normalization in the Symmetric Phase . . . . . . . . . . . . . 229Shiro Takagi, Yuki Yoshida, and Masato Okada

Regularized Pooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241Takato Otsuzuki, Hideaki Hayashi, Yuchen Zheng, and Seiichi Uchida

Reinforcement Learning I

Deep Recurrent Deterministic Policy Gradient for Physical Control . . . . . . . . 257Lei Zhang, Shuai Han, Zhiruo Zhang, Lefan Li, and Shuai Lü

Exploration via Progress-Driven Intrinsic Rewards . . . . . . . . . . . . . . . . . . . 269Nicolas Bougie and Ryutaro Ichise

xxii Contents – Part II

An Improved Reinforcement Learning Based Heuristic DynamicProgramming Algorithm for Model-Free Optimal Control . . . . . . . . . . . . . . 282

Jia Li, Zhaolin Yuan, and Xiaojuan Ban

PBCS: Efficient Exploration and Exploitation Using a Synergy BetweenReinforcement Learning and Motion Planning . . . . . . . . . . . . . . . . . . . . . . 295

Guillaume Matheron, Nicolas Perrin, and Olivier Sigaud

Understanding Failures of Deterministic Actor-Critic with ContinuousAction Spaces and Sparse Rewards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308

Guillaume Matheron, Nicolas Perrin, and Olivier Sigaud

Reinforcement Learning II

GAN-Based Planning Model in Deep Reinforcement Learning . . . . . . . . . . . 323Song Chen, Junpeng Jiang, Xiaofang Zhang, Jinjin Wu,and Gongzheng Lu

Guided Reinforcement Learning via Sequence Learning . . . . . . . . . . . . . . . . 335Rajkumar Ramamurthy, Rafet Sifa, Max Lübbering,and Christian Bauckhage

Neural Machine Translation Based on Improved Actor-Critic Method . . . . . . 346Ziyue Guo, Hongxu Hou, Nier Wu, and Shuo Sun

Neural Machine Translation Based on Prioritized Experience Replay . . . . . . . 358Shuo Sun, Hongxu Hou, Nier Wu, and Ziyue Guo

Improving Multi-agent Reinforcement Learning with ImperfectHuman Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369

Xiaoxu Han, Hongyao Tang, Yuan Li, Guang Kou, and Leilei Liu

Reinforcement Learning III

Adaptive Skill Acquisition in Hierarchical Reinforcement Learning . . . . . . . . 383Juraj Holas and Igor Farkaš

Social Navigation with Human Empowerment Driven DeepReinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395

Tessa van der Heiden, Florian Mirus, and Herke van Hoof

Curious Hierarchical Actor-Critic Reinforcement Learning . . . . . . . . . . . . . . 408Frank Röder, Manfred Eppe, Phuong D. H. Nguyen, and Stefan Wermter

Policy Entropy for Out-of-Distribution Classification . . . . . . . . . . . . . . . . . . 420Andreas Sedlmeier, Robert Müller, Steffen Illium,and Claudia Linnhoff-Popien

Contents – Part II xxiii

Reservoir Computing

Analysis of Reservoir Structure Contributing to Robustness AgainstStructural Failure of Liquid State Machine . . . . . . . . . . . . . . . . . . . . . . . . . 435

Yuta Okumura and Naoki Wakamiya

Quantifying Robustness and Capacity of Reservoir Computerswith Consistency Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447

Thomas Lymburn, Thomas Jüngling, and Michael Small

Two-Step FORCE Learning Algorithm for Fast Convergencein Reservoir Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

Hiroto Tamura and Gouhei Tanaka

Morphological Computation of Skin Focusing on Fingerprint Structure . . . . . 470Akane Musha, Manabu Daihara, Hiroki Shigemune,and Hideyuki Sawada

Time Series Clustering with Deep Reservoir Computing . . . . . . . . . . . . . . . 482Miguel Atencia, Claudio Gallicchio, Gonzalo Joya, and Alessio Micheli

ReservoirPy: An Efficient and User-Friendly Library to DesignEcho State Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494

Nathan Trouvain, Luca Pedrelli, Thanh Trung Dinh, and Xavier Hinaut

Robotics and Neural Models of Perception and Action

Adaptive, Neural Robot Control – Path Planning on 3D Spiking NeuralNetworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509

Lea Steffen, Artur Liebert, Stefan Ulbrich, Arne Roennau,and Rüdiger Dillmannn

CABIN: A Novel Cooperative Attention Based Location PredictionNetwork Using Internal-External Trajectory Dependencies . . . . . . . . . . . . . . 521

Tangwen Qian, Fei Wang, Yongjun Xu, Yu Jiang, Tao Sun, and Yong Yu

Neuro-Genetic Visuomotor Architecture for Robotic Grasping . . . . . . . . . . . 533Matthias Kerzel, Josua Spisak, Erik Strahl, and Stefan Wermter

From Geometries to Contact Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546Martin Meier, Robert Haschke, and Helge J. Ritter

Sentiment Classification

Structural Position Network for Aspect-Based Sentiment Classification . . . . . 559Pu Song, Wei Jiang, Fuqing Zhu, Yan Zhou, Jizhong Han,and Songlin Hu

xxiv Contents – Part II

Cross-Domain Sentiment Classification Using Topic Attentionand Dual-Task Adversarial Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571

Kwun-Ping Lai, Jackie Chun-Sing Ho, and Wai Lam

Data Augmentation for Sentiment Analysis in English –

The Online Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584Michał Jungiewicz and Aleksander Smywiński-Pohl

Spiking Neural Networks I

Dendritic Computation in a Point Neuron Model. . . . . . . . . . . . . . . . . . . . . 599Alexander Vandesompele, Francis Wyffels, and Joni Dambre

Benchmarking Deep Spiking Neural Networkson Neuromorphic Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 610

Christoph Ostrau, Jonas Homburg, Christian Klarhorst, Michael Thies,and Ulrich Rückert

Unsupervised Learning of Spatio-Temporal Receptive Fieldsfrom an Event-Based Vision Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622

Thomas Barbier, Céline Teulière, and Jochen Triesch

Spike-Train Level Unsupervised Learning Algorithm for Deep SpikingBelief Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 634

Xianghong Lin and Pangao Du

Spiking Neural Networks II

Modelling Neuromodulated Information Flow and Energetic Consumptionat Thalamic Relay Synapses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649

Mireille Conrad and Renaud B. Jolivet

Learning Precise Spike Timings with Eligibility Traces . . . . . . . . . . . . . . . . 659Manuel Traub, Martin V. Butz, R. Harald Baayen, and Sebastian Otte

Meta-STDP Rule Stabilizes Synaptic Weights Under in Vivo-like OngoingSpontaneous Activity in a Computational Model of CA1 Pyramidal Cell . . . . 670

Matúš Tomko, Peter Jedlička, and L’ubica Beňušková

Adaptive Chemotaxis for Improved Contour Tracking Using SpikingNeural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 681

Shashwat Shukla, Rohan Pathak, Vivek Saraswat, and Udayan Ganguly

Contents – Part II xxv

Text Understanding I

Mental Imagery-Driven Neural Network to Enhance Representationfor Implicit Discourse Relation Recognition . . . . . . . . . . . . . . . . . . . . . . . . 695

Jian Wang, Ruifang He, Fengyu Guo, and Yugui Han

Adaptive Convolution Kernel for Text Classificationvia Multi-channel Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 708

Cheng Wang and Xiaoyan Fan

Text Generation in Discrete Space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721Ting Hu and Christoph Meinel

Short Text Processing for Analyzing User Portraits:A Dynamic Combination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733

Zhengping Ding, Chen Yan, Chunli Liu, Jianrui Ji, and Yezheng Liu

A Hierarchical Fine-Tuning Approach Based on Joint Embedding of Wordsand Parent Categories for Hierarchical Multi-label Text Classification . . . . . . 746

Yinglong Ma, Jingpeng Zhao, and Beihong Jin

Text Understanding II

Boosting Tricks for Word Mover’s Distance. . . . . . . . . . . . . . . . . . . . . . . . 761Konstantinos Skianis, Fragkiskos D. Malliaros, Nikolaos Tziortziotis,and Michalis Vazirgiannis

Embedding Compression with Right Triangle Similarity Transformations . . . . 773Haohao Song, Dongsheng Zou, Lei Hu, and Jieying Yuan

Neural Networks for Detecting Irrelevant Questions During VisualQuestion Answering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786

Mengdi Li, Cornelius Weber, and Stefan Wermter

F-Measure Optimisation and Label Regularisation for Energy-Based NeuralDialogue State Tracking Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 798

Anh Duong Trinh, Robert J. Ross, and John D. Kelleher

Unsupervised Learning

Unsupervised Change Detection Using Joint Autoencoders for Age-RelatedMacular Degeneration Progression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813

Guillaume Dupont, Ekaterina Kalinicheva, Jérémie Sublime,Florence Rossant, and Michel Pâques

A Fast Algorithm to Find Best Matching Units in Self-Organizing Maps . . . . 825Yann Bernard, Nicolas Hueber, and Bernard Girau

xxvi Contents – Part II

Tumor Characterization Using Unsupervised Learning of MathematicalRelations Within Breast Cancer Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 838

Cristian Axenie and Daria Kurz

Balanced SAM-kNN: Online Learning with Heterogeneous Driftand Imbalanced Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 850

Valerie Vaquet and Barbara Hammer

A Rigorous Link Between Self-Organizing Maps and GaussianMixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 863

Alexander Gepperth and Benedikt Pfülb

Collaborative Clustering Through Optimal Transport . . . . . . . . . . . . . . . . . . 873Fatima Ezzahraa Ben Bouazza, Younès Bennani, Guénaël Cabanes,and Abdelfettah Touzani

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887

Contents – Part II xxvii