Lecture Notes in Computer Science 10597 Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen Editorial Board David Hutchison Lancaster University, Lancaster, UK Takeo Kanade Carnegie Mellon University, Pittsburgh, PA, USA Josef Kittler University of Surrey, Guildford, UK Jon M. Kleinberg Cornell University, Ithaca, NY, USA Friedemann Mattern ETH Zurich, Zurich, Switzerland John C. Mitchell Stanford University, Stanford, CA, USA Moni Naor Weizmann Institute of Science, Rehovot, Israel C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen TU Dortmund University, Dortmund, Germany Demetri Terzopoulos University of California, Los Angeles, CA, USA Doug Tygar University of California, Berkeley, CA, USA Gerhard Weikum Max Planck Institute for Informatics, Saarbrücken, Germany
26
Embed
Lecture Notes in Computer Science 10597 - Springer978-3-319-69900-4/1.pdf · Lecture Notes in Computer Science 10597 Commenced Publication in 1973 Founding and Former Series Editors:
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Lecture Notes in Computer Science 10597
Commenced Publication in 1973Founding and Former Series Editors:Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David HutchisonLancaster University, Lancaster, UK
Takeo KanadeCarnegie Mellon University, Pittsburgh, PA, USA
Josef KittlerUniversity of Surrey, Guildford, UK
Jon M. KleinbergCornell University, Ithaca, NY, USA
Friedemann MatternETH Zurich, Zurich, Switzerland
John C. MitchellStanford University, Stanford, CA, USA
Moni NaorWeizmann Institute of Science, Rehovot, Israel
C. Pandu RanganIndian Institute of Technology, Madras, India
Bernhard SteffenTU Dortmund University, Dortmund, Germany
Demetri TerzopoulosUniversity of California, Los Angeles, CA, USA
Doug TygarUniversity of California, Berkeley, CA, USA
Gerhard WeikumMax Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7412
This Springer imprint is published by Springer NatureThe registered company is Springer International Publishing AGThe registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
It is a great pleasure to introduce you all to the proceedings of the 7th InternationalConference on Pattern Recognition and Machine Intelligence (PReMI 2017), held atthe Indian Statistical Institute (ISI), Kolkata, India, during December 5–8, 2017. Theobjective of the conference is to introduce to the community the most recentadvancements in research in the domain of pattern recognition and machine intelli-gence. Our goal is to encourage academic and industrial collaboration in all relatedfields in machine learning involving scientists, engineers, professionals, researchers,and students from India and abroad. The conference is held biennially to make it anideal platform for researcher all over the world to come and share their views andexperiences. This was the seventh edition in this series, being held in the year markingthe 125th birthday of late Prof. Prasanta Chandra Mahalanobis.
Professor Mahalanobis was the founder of the Indian Statistical Institute and thefather of modern statistics in India. As researchers in pattern recognition and machinelearning we are immensely indebted to him. He was instrumental in inspiring thedesign of the first analog computer in India in 1953. He brought to ISI the first digitalcomputer to India in the year 1955. As a mark of our respect to this monumentalpersonality, we organized a Special Session on “Celebration of 125th Birth Anniver-sary of Professor P.C. Mahalanobis” at PReMI 2017.
The conference comprised several keynote and invited lecturers delivered by emi-nent and distinguished researchers from around the world. Both the invited and thetechnical sessions featured interesting lectures in classic and contemporary aspects ofmachina intelligence. The topics range from deep learning and Internet of Things(IoT) to computer vision and big data analytics. There were two exclusive sessions on“Deep Learning” and “Spatial Data Science and Engineering” Like previous editions,PReMI 2017 had a very good response in terms of paper submissions. Altogether therewere 293 submissions from about 15 countries spanning three continents. Each paperwas critically reviewed by experts in the field, after which 85 papers (29% acceptancerate) were accepted for inclusion in these proceedings. The accepted papers are dividedinto ten groups, although there could be some overlap. Articles written by the keynoteand invited speakers are also included in the proceedings (mostly abstracts).
We wish to express our appreciation to the Program Committee and TechnicalReview Committee members, who worked hard to ensure the quality of the contri-butions of this volume. We are thankful to the editors of the journals FundamentaInformaticae and Applied Soft Computing for kindly agreeing to publish the extendedversions of some of the selected papers in their esteemed journals. We also take thisopportunity to thank Professors Vineet Bafna, Andrzej Skowron, Farzin Deravi,Upinder S. Bhalla, Uday B. Desai, Soumen Chakraborti, Ambarish Ghosh, ParthaPratim Majumder, Probal Chaudhuri, Subhasis Chaudhuri, David Zhang, and ShalabhBhatnagar for accepting our invitation to deliver keynote, invited, and special lecturesduring the conference. We gratefully acknowledge Alfred Hofmann of Springer for his
co-operation in the publication of the PReMI 2017 proceedings in the LNCS series, asdone for the previous editions. We would like to thank all the organizations who eitherendorsed or sponsored this conference technically or financially. We are grateful toEasyChair for providing us with a wonderful platform for conducting the entire processof paper review. Last but not the least, we take this opportunity to thank all thecontributors for their enthusiastic response, without which no conference can ever besuccessful.
While preparing the proceedings we mourned the sad demise of Professor Lotfi A.Zadeh, the founder of fuzzy mathematics, an imperative part of contemporary machinelearning. He was on the advisory board of PReMI ever since its inception in 2005,including the present edition. Our institute honored him with a doctor honoris causa in2006 during its annual convocation. We express our deep condolences to his familyand all his friends and colleagues. It is a great loss to the pattern recognition and softcomputing/computational intelligence community.
Our best wishes to all the participants of PReMI 2017 conference. May this volume,which contains the papers presented at PReMI 2017 prove to be a valuable source ofreference for ongoing and future research work.
December 2017 B. Uma ShankarKuntal Ghosh
Deba Prasad MandalShubhra Sankar Ray
David ZhangSankar K. Pal
VI Preface
Message from the General Chair
PReMI, the biennial International Conference on Pattern Recognition and MachineIntelligence, returned to Kolkata, the City of Joy, after its sixth edition in Warsaw,Poland, in June/July 2015! I am delighted that the seventh edition (PReMI 2017) washeld in the year that marks the 125th birthday of late Prof. Prasanta Chandra Maha-lanobis, the founder of our Indian Statistical Institute.
Like earlier versions, PReMI 2017 had a nice mixture of keynote and invitedspeeches, and quality research papers using both classic and modern computingparadigms, covering different facets of pattern recognition and machine intelligencewith real-life applications. Apart from classic topics, special emphasis was given tocontemporary research areas such as big data analytics, deep learning, Internet ofThings, and computer vision through both regular and special sessions. Somepost-conference special issues will be published as done in the past. All these makePReMI 2017 an ideal state-of-the-art platform for researchers and practitioners toexchange ideas and enrich their knowledge.
I thank all the participants, speakers, reviewers, and members of various committeesfor making this event a grand success. My thanks are also due to the sponsors for theirsupport, and Springer for publishing the PReMI proceedings, since its first edition in2005, in the prestigious LNCS series.
I trust, the participants had an academically fruitful and enjoyable stay in Kolkata.
December 2017 Sankar K. Pal
Organization
PReMI 2017 was organized by the Machine Intelligence Unit, Indian StatisticalInstitute (ISI) in Kolkata during December 5–8, 2017.
Conference Committee
Patron
SanghamitraBandyopadhyay
ISI, Kolkata, India
General Chair
Sankar K. Pal ISI, Kolkata, India
Program Chairs
David Zhang PolyU, Hong Kong, SAR ChinaKuntal Ghosh ISI, Kolkata, IndiaB. Uma Shankar ISI, Kolkata, India
Organizing Chairs
Deba Prasad Mandal ISI, Kolkata, IndiaShubhra Sankar Ray ISI, Kolkata, India
Special Session Chairs
Ashish Ghosh ISI, Kolkata, IndiaFarid Melgani University of Trento, ItalySambhunath Biswas TIU, Kolkata, IndiaAlfredo Petrosino University of Naples, ItalySoumya K. Ghosh IIT Kharagpur, India
Malay K. Kundu ISI, Kolkata, IndiaC.A. Murthy ISI, Kolkata, IndiaSantanu Chaudhury CEERI, Pilani, India
International Liaisons
Sergei O. Kuznetsov HSE, Moscow, RussiaMarzena Kryszkiewicz WUT, PolandSimon C.K. Shiu PolyU, Hong Kong, SAR China
Advisory Committee
Lofti A. Zadeh, USAC.R. Rao, USAAnil K. Jain, USAJosef Kittler, UKLaveen N. Kanal, USAB.L. Deekshatulu, IndiaDwijesh Dutta Majumder, IndiaAndrzej Skowron, PolandRama Chellappa, USAWitold Pedrycz, CanadaDavid W. Aha, USAGabriella Sanniti di Baja, ItalyB. Yegnanarayana, IndiaShun-ichi Amari, JapanJayaram Udupa, USAJiming Liu, Hong Kong, SAR ChinaRonald Yager, USANing Zhong, JapanTharram Dillon, AustraliaHenryk Rybinski, Poland
Program Committee
Jayadeva Indian Institute of Technology Delhi, IndiaTinku Acharya Videonetics Technology Pvt. Ltd., IndiaMd. Atiqur Rahman Ahad University of Dhaka, BangladeshMohua Banerjee Indian Institute of Technology Kanpur, IndiaJayanta Basak NetApp, IndiaSmarajit Bose Indian Statistical Institute, IndiaRoberto M. Cesar USP, BrazilGoutam Chakraborty Iwate Prefectural University, JapanBhabatosh Chanda Indian Statistical Institute, IndiaSubhasis Chaudhuri Indian Institute of Technology Bombay, IndiaSung-Bae Cho Yonsei University, South KoreaPartha Pratim Das Indian Institute of Technology Kharagpur, IndiaSukhendu Das Indian Institute of Technology Madras, IndiaDipankar Dasgupta The University of Memphis, USARajat K. De Indian Statistical Institute, India
X Organization
Farzin Deravi University of Kent, UKShaikh A. Fattah BUET, BangladeshPaolo Gamba University of Pavia, ItalyJoydeep Ghosh University of Texas, USAMark Girolami University of Warwick, UKPhalguni Gupta NITTTR Kolkata, IndiaLarry Hall University of South Florida, USAFrancisco Herrera University of Granada, SpainQinghua Hu Tianjin University, ChinaC.V. Jawahar IIIT Hyderabad, IndiaJohn Kerekes Rochester Institute of Technology, USARavi Kothari IBM Research, IndiaPawan Lingras Saint Mary’s University, USAPradipta Maji Indian Statistical Institute, IndiaFrancesco Masulli University of Genoa, ItalyPabitra Mitra Indian Institute of Technology Kharagpur, IndiaJayanta Mukherjee Indian Institute of Technology Kharagpur, IndiaM.N. Murty Indian Institute of Science, IndiaY. Narahari Indian Institute of Science, IndiaB.L. Narayan Yahoo! Labs, USANasser Nasrabadi West Virginia University, USANikhil Rajan Pal Indian Statistical Institute, IndiaAngel P. Del Pobil Universitat Jaume I, SpainAmit K. Roy-Chowdhury University of California, USAPunam Saha University of Iowa, USAP.S. Sastry Indian Institute of Science, IndiaFaisal Shafait National University of Sciences and Technology,
PakistanSitabhra Sinha The Institute of Mathematical Sciences, IndiaDominik Slezak University of Warsaw, PolandBrijesh Verma Central Queensland University, AustraliaDianhui Wang La Trobe University, Australia
International Association for Pattern Recognition (IAPR)
Technical Co-sponsor
IEEE Kolkata Section
Other Sponsors
Center for Soft Computing Research: A National Facility, ISI, KolkataWeb Intelligence Consortium (WIC)International Rough Set Society (IRSS)INAE Kolkata ChapterWorld Federation on Soft Computing (WFSC)Springer International Publishing
XII Organization
Abstracts of Invited Talks
Interactive Granular Computingin Data Science
Andrzej Skowron1, 2
1 Faculty of Mathematics, Computer Science and Mechanics,University of Warsaw, [email protected]
2 Systems Research Institute, Polish Academy of Sciences
We discuss Interactive Granular Computing (IGrC) as the basis of a Data Sciencecomputing model. IGrC binds together and brings a synchronous cooperation amongthe following four basic concepts of Artificial Intelligence: language, reasoning, per-ception, and action. This, together with information granulation, helps agents to dealwith many complex tasks of perceiving or transforming compound abstract andphysical objects (e.g., in the context of complex spatio-temporal space). One shouldconsider that in Data Science agents collecting data have control over the dataacquisition, i.e., they are deciding say which data, using which sources, at what time,and why should be collected.
Basic objects in IGrC are complex granules (c-granules or granules, for short).They are grounded in the physical reality and are, in particular, responsible for gen-eration of the networks of information systems (data tables) through interactions withthe configurations of physical objects. Development of a particular network of infor-mation systems is guided by the need to learn the relevant computational buildingblocks that are necessary for perception, using the formulation by Leslie Valiant.Among these blocks, often learned hierarchically, one can distinguish patterns, clustersor classifiers. The computational building blocks are used by agents, e.g., forapproximation of conditions responsible for initiating actions or plans. Agents per-forming computations based on interaction with the physical environment learn newc-granules, in particular, in the form of interaction rules, representing knowledge notknown a priori by agents. These new c-granules are used not only for construction ofcompound abstract objects but also of compound physical objects, e.g., sensors com-posed out of more primitive sensors. Learning of interaction rules also supports thecontrol of agents, in particular the self-organized distributed control. Numerous tasks ofagents may be classified as control tasks performed by agents aiming at achieving thehigh quality computational trajectories of configurations of c-granules relative to theconsidered quality measures over the trajectories.
Reasoning supporting agents in searching for solutions of their tasks is based onadaptive judgment, an important component of IGrC. Methods based on adaptivejudgment allow agents to construct from given configurations of their c-granules newones. These new configurations of c-granules should be constructed taking into accountthe needs of agents realized through interactions with the environment. Here, new
challenges are related to developing strategies for predicting and controlling behaviorsof agents. We propose to investigate these challenges using the IGrC framework withadaptive judgment used for controlling of computations performed on c-granules. Forexample, adaptive judgment is used in adaptive learning of rough set based approxi-mations of complex vague concepts evolving with time. It is also used in the riskmanagement of granular computations, carried out by agents, toward achieving theagent needs.
XVI A. Skowron
Identifying the Favored Allelein a Selective Sweep
Vineet Bafna
Computer Science and Engineering University of California,San Diego, USA
Abstract. Selection is a dominant force in evolution. Mutations arising at ran-dom might favor individuals in a specific environmental niche, and populationsadapt by rapidly increasing the frequency of individuals carrying the favoredmutations. The selection process results in distinct patterns (a signature) of allelefrequencies and haplotype structures that can be exploited to identify the genesresponding to selection pressure. A study of selection signals in humans has ledto molecular insight into the evolution of many natural traits such as skin andeye color, as also adaptation to extreme environments.
Computational methods that scan population genomics data to identifysignatures of selective sweep have been actively developed, but mostly do notidentify the specific mutation favored by the selective sweep. In this talk, wedescribe an approach that uses population genetics and machine learning tech-niques to pin-point the favored mutation, even when the signature of selectionextends to 5Mbp. Our method, iSAFE, was tested extensively on simulated dataand 22 known sweeps in human populations using the 1000 genome project datawith some evidence for the favored mutation. iSAFE ranked the candidatemutation among the top 15 (out of * 21,000 candidates) in 14 of the 22 loci,and identified previously unreported mutations as favored the 5 regions.
Sequence Recognition as a SubcellularComputational Primitive in Neural Function
Upinder S. Bhalla
National Centre for Biological Sciences (NCBS), Bangalore, [email protected]
Abstract. Many sensory, motor, and cognitive processes involve sequenceswith complex hierarchical structures. In computational neuroscience these havetypically been modeled as arising from network computation. We have analyzedhow such computations may arise instead from subcellular reaction-diffusionprocesses on small (*30 micron) segments of neuronal dendrites. This for-mulation vastly increases the potential computational capacity of neuronalnetworks. We consider some possible mappings of subcellular sequence com-putation to the structure of deep learning networks. This is interesting because itprovides for very compact and efficient biological implementations ofLSTM-like networks. We speculate that there may be a parallel between someof the computational principles of engineered networks and the hippocampal-entorhinal cortex loop.
An Incremental Fast Policy SearchUsing a Single Sample Path
Ajin George Joseph and Shalabh Bhatnagar
Indian Institute of Science, Bangalore, India{ajin,shalabh}@iisc.ac.in
Abstract. In this paper, we consider the control problem in a reinforcementlearning setting with large state and action spaces. The control problem mostcommonly addressed in the contemporary literature is to find an optimal policywhich optimizes the long run c -discounted transition costs, where c 2 0; 1½ Þ .They also assume access to a generative model/simulator of the underlyingMDP with the hidden premise that realization of the system dynamics of theMDP for arbitrary policies in the form of sample paths can be obtained with easefrom the model. In this paper, we consider a cost function which is theexpectation of a approximate value function w.r.t. the steady state distributionof the Markov chain induced by the policy, without having access to the gen-erative model. We assume that a single sample path generated using a priorichosen behaviour policy is made available. In this information restricted setting,we solve the generalized control problem using the incremental cross entropymethod. The proposed algorithm is shown to converge to the solution which isglobally optimal relative to the behaviour policy.
Biometric Counter-Spoofing for MobileDevices Using Gaze Information
Abstract.With the rise in the use of biometric authentication on mobile devices,it is important to address the security vulnerability of spoofing attacks where anattacker using an artefact representing the biometric features of a genuine userattempts to subvert the system. In this paper, techniques for presentation attackdetection are presented using gaze information with a focus on their applicabilityfor use on mobile devices. Novel features that rely on directing the gaze of theuser and establishing its behaviour are explored for detecting spoofing attempts.The attack scenarios considered in this work include the use of projected photos,2D and 3D masks. The proposed features and the systems based on them wereextensively evaluated using data captured from volunteers performing genuineand spoofing attempts. The results of the evaluations indicate that gaze-basedfeatures have the potential for discriminating between genuine attempts andimposter attacks on mobile devices.
Abhijit Dasgupta, Losiana Nayak, Ritankar Das, Debasis Basu,Preetam Chandra, and Rajat K. De
Selection of Relevant Electrodes Based on Temporal Similarityfor Classification of Motor Imagery Tasks . . . . . . . . . . . . . . . . . . . . . . . . . 96
Jyoti Singh Kirar, Ayesha Choudhary, and R.K. Agrawal
Automated Measurement of Translational Margins and Rotational Shiftsin Pelvic Structures Using CBCT Images of Rectal Cancer Patients. . . . . . . . 103
A Machine Learning Inspired Approach for Detection, Recognitionand Tracking of Moving Objects from Real-Time Video . . . . . . . . . . . . . . . 170
Anit Chakrabory and Sayandip Dutta
Does Rotation Influence the Estimated Contour Length of a Digital Object? . . . 179Sabyasachi Mukherjee, Oishila Bandyopadhyay, Arindam Biswas,and Bhargab B. Bhattacharya
Abnormal Crowd Behavior Detection Based on CombinedApproach of Energy Model and Threshold . . . . . . . . . . . . . . . . . . . . . . . . . 187
Music-Induced Emotion Classification from the Prefrontal Hemodynamics . . . 289Pallabi Samanta, Diptendu Bhattacharya, Amiyangshu De, Lidia Ghosh,and Amit Konar
Analysis of Causal Interactions and Predictive Modelling of FinancialMarkets Using Econometric Methods, Maximal Overlap Discrete WaveletTransformation and Machine Learning: A Study in Asian Context . . . . . . . . 664