table of contents - ncbr€¦ · prof. krzysztof krawiec deep feature engineering for segmentation,...
TRANSCRIPT
1
Table of Contents Agenda of the Poland-Taiwan Seminar on Artificial Intelligence ............................................................. 3
Piotr Sankowski ....................................................................................................................................... 7
Youn-Long Lin ......................................................................................................................................... 8
Abstract ................................................................................................................................................ 9
Tsung-Yi Ho ........................................................................................................................................... 10
Dominik Ślęzak ...................................................................................................................................... 11
Abstract .............................................................................................................................................. 12
Chang-Shing Lee ................................................................................................................................... 13
Abstract .............................................................................................................................................. 14
Krzysztof Krawiec .................................................................................................................................. 15
Abstract .............................................................................................................................................. 15
Hsun-Ping Hsieh .................................................................................................................................... 16
Abstract .............................................................................................................................................. 16
Jakub Nalepa ......................................................................................................................................... 17
Abstract .............................................................................................................................................. 17
Jason S. Chang ..................................................................................................................................... 19
Abstract .............................................................................................................................................. 19
Grzegorz J. Nalepa ................................................................................................................................ 21
Abstract .............................................................................................................................................. 21
Ming-Feng Tsai ...................................................................................................................................... 23
Abstract .............................................................................................................................................. 23
Radosław Gwarek ................................................................................................................................. 24
Abstract .............................................................................................................................................. 25
Piotr Babieno ......................................................................................................................................... 26
Abstract .............................................................................................................................................. 26
Sheng-Jyh Wang ................................................................................................................................... 28
Abstract .............................................................................................................................................. 28
Cezary Dołęga ....................................................................................................................................... 29
Abstract .............................................................................................................................................. 30
Winston Hsu .......................................................................................................................................... 31
Abstract .............................................................................................................................................. 31
Bartosz Wilczyński ................................................................................................................................. 32
Abstract .............................................................................................................................................. 32
Notes ..................................................................................................................................................... 33
2
Ladies and gentlemen,
It is my great privilege to present to you the Booklet of Poland – Taiwan Seminar on Artificial Intelligence.
The National Centre for Research and Development (NCBR) was established in 2007. This year the Centre celebrates the 10th anniversary of its presence in the national ecosystem. With our programmes we have managed to respond to the needs of entrepreneurs seeking assistance in their efforts to become more innovative. We have created a strong support system for science institutions, entrepreneurs and consortia that are willing to join forces in transforming Polish economy.
We are proud to observe more and more beneficiaries of NCBR programmes who make their mark on the national or global market with newly developed, advanced technologies, implementing innovations and challenging global competitors.
International activity is one of the crucial components of our programme portfolio, in line with the Foreign Expansion pillar of Poland’s economic development. NCBR is involved in a variety of programmes executed with sister agencies abroad. Initiated in 2012 by National Centre for Research and Development and National Science Council, the bilateral programme with Taiwan has proved to be a very successful story in offering financial support to Polish and Taiwanese research teams. To this date NCBR and MOST selected 28 projects for co-financing. The overall budget of these projects amounts to 2,9 M USD. The annual calls for proposals include the technologies which correspond to priority areas of the two countries: neuroscience, energy efficiency, materials science, agricultural biotechnology and biocatalysis, ICT and artificial intelligence.
In order to spread the idea of bilateral calls for research projects among potential stakeholders in both countries, the two funding parties decided to launch seminars on an annual basis. This year’s edition will be the 6th meeting of distinguished Professors and innovation-oriented entrepreneurs where they present the most promising ideas for future collaboration under the umbrella of joint calls. The Artificial Intelligence Seminar takes place in Warsaw while every second year the seminar is organised in Taiwan. It engages representatives of institutions and companies dealing with science and innovations related to the topic of the seminar.
I am very much pleased to convey to all participants of the Poland – Taiwan Seminar on Artificial Intelligence the wishes of interesting presentations which shall inspire to create and implement ambitious projects under Polish – Taiwanese collaboration. NCBR will continue to support the funding scheme in the future and endeavour to promote the calls thoroughly. Our goal in all bilateral programmes is to help Polish teams become more competitive on the international research stage. We also want to attract Polish researchers who developed their scientific careers abroad as well as foreign investigators to use world-class research infrastructure available at Polish universities and technology centres.
Let this seminar be a big step forward in bilateral cooperation in artificial intelligence!
Prof. Maciej Chorowski
3
Agenda of the Poland-Taiwan Seminar on Artificial Intelligence
October 18, 2017, Venue: Baltic Ballroom I, Warsaw Marriott Hotel,
Al. Jerozolimskie 65/79, 00-697 Warsaw
8.30 – 9.00
Registration and coffee
9.00 – 9.15
Welcome speech
Prof. Aleksander Nawrat
Deputy Director of the National Centre for Research and Development
Ambassador Weber V. B. Shih
Representative and Head of the Taipei Economic and Cultural Office in Poland
9.15 – 9.20
Introduction – coordinator
Prof. Piotr Sankowski
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw
9.20 – 9.40
Prof. Youn-Long Lin
Introduction to AI Innovation Center in Taiwan
Department of Computer Science, National Tsing Hua University
9.40 – 10.00
Prof. Dominik Ślęzak
Toward Approximate Intelligence – Lessons Learnt from Development of KDD Methods
and Approximate Query Engines
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw
10.00 – 10.20
Prof. Chang-Shing Lee
Intelligent Agent for Human and Machine Co-Learningon Game of Go
Department of Computer Science and Information Engineering, National University of Tainan
4
10.20 – 10.40
Prof. Krzysztof Krawiec
Deep Feature Engineering for Segmentation, Anomaly Detection
and Diagnosing in Medical Imaging
Faculty of Computing, Poznan University of Technology
10.40 – 11.05
Coffee break
11.05 – 11.25
Prof. Hsun-Ping Hsieh
Mining Geo-Social Media and Urban Informatics
for Smart Environmental and Business Applications
Department of Electrical Engineering, National Cheng Kung University
11.25 – 11.45
Prof. Jakub Nalepa
Towards Automatic Design of Deep Neural Networks Through Evolutionary Computation
Department of Automatic Control, Electronics and Computer Science,
Silesian University of Technology
11.45 – 12.05
Prof. Jason S. Chang
Inducing Lexical Grammar from Monolingual and Bilingual Corpora
for Computer Assisted Language Learning and Translation
Department of Computer Science, National Tsing Hua University
12.05 – 12.25
Prof. Grzegorz J. Nalepa
Context, Autonomy, Affect. Methodological and Ethical Challenges of Today's AI
Department of Applied Computer Science, AGH University of Science and Technology
12.25 – 13.25
Lunch break
5
13.25 – 13.45
Prof. Ming-Feng Tsai
Textual Data Analytics in Finance
Department of Computer Science, National Chengchi University
13.45 – 14.05
Radosław Gwarek
Meta AI – The artificial mind
11bit studios SA
14.05 – 14.25
Piotr Babieno
Hidden Horror and Catharsis 2.0
CEO Bloober Team SA
14.25 – 14.45
Coffee break
14.45 – 15.05
Prof. Sheng-Jyh Wang
An Efficient Sequential Partitioning Method for Clustering and Manifold Learning
Department of Electronic Engineering, National Chiao Tung University
15.05 – 15.25
Cezary Dołęga
NeuroCar – Applications of artificial intelligence in Intelligent Transport Systems:
traffic safety, intelligent infrastructure and autonomous vehicles
Neurosoft Sp. z o.o.
15.25 – 15.45
Prof. Winston Hsu
Learning from Multimodal Data Streams
Department of Computer Science and Information Engineering, National Taiwan University
6
15.45 – 16.05
Prof. Bartosz Wilczyński
Machine learning at work in modern functional genomics
Faculty of Mathemathics, Informatics and Mechanics, University of Warsaw
16.05 – 16.30
Conclusions and expectations for the future – moderator and all (general discussion)
Closing remarks – moderator and coordinators
7
Piotr Sankowski
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw
e-mail: [email protected]
phone number: +48 22 55 44 427
Piotr Sankowski is Professor of Computer Science at Warsaw University. He received
a habilitation in computer science in 2009 and a PhD in computer science in 2005, both from
the University of Warsaw, where he was working on combinatorial optimization problems, with
special emphasis on dynamic computations and stochastic properties of data. He also received
a PhD in physics in 2009 from the Polish Academy of Sciences, where he was working on solid
state theory. In 2010, he received an ERC Starting Independent Researcher Grant, and in
2015 an ERC Proof of Concept grant. He is a co-founder of MIM Solutions – a spin-off company
that aims to commercialize algorithmic market modelling.
8
Youn-Long Lin
Department of Computer Science, National Tsing Hua University
Program Office Director, Taiwan AI Innovation Research
e-mail: [email protected]
phone number: +886-975-123-895
Youn-Long Lin is a Tsing Hua Chair Professor of Computer Science of National Tsing Hua
University. He was born in Yun-Lin, Taiwan. He received his B.S. degree in electronics
engineering from National Taiwan University of Science and Technology (formerly, National
Taiwan Institute of Technology), Taipei, Taiwan, in 1982, and his Ph.D. in computer science
from the University of Illinois, Urbana-Champaign, IL, U.S.A. in 1987. Upon his graduation,
he joined National Tsing Hua University, Hsin-Chu, Taiwan, where he established the THEDA
Group (Tsing Hua EDA), served as Director of the University Computer and Communication
Center, Chairman of the Department of Computer Science, Secretariat General of the
University, Chief Librarian of the University Library, and the Dean of Research & Development.
He is also an adjunct professor of Peking University, Beijing and a guest professor of Waseda
University, Japan.
Between 1977-1979, Dr. Lin was in military service with Chung Cheng Armed Forces
Preparatory School as a Second Lieutenant. In 1979-1980, he was with Taiwan Cement Corp.
as a field instrument engineer. In 1981-1983, he was with the Atomic Power Division of Taiwan
Power Corp. as a mini-computer system administrator and database AP developer.
In 1984-1987, he was a research assistant with the Department of Computer Science,
University of Illinois, Urbana-Champaign, IL, USA. In 1998-1999, on sabbatical leave from
Tsing Hua, he co-founded Global UniChip Corp., a SOC Design Foundry, serving as its Chief
Technical Advisor. Between 2001 and 2003, he took a leave-of-absence from Tsing Hua
working for UniChip as its Chief Technical Officer and Executive Vice President.
Professor Lin's primary research interest is in computer-aided design (CAD) of very large-scale
integrated circuits (VLSI) with emphasis on physical design automation, high-level synthesis,
and VLSI design for video coding. He co-authored the book “High Level Synthesis –
Introduction to Chip and System Design.” He edited the book “Essential Issues in SOC
Design.” He also spend great effort in promoting the VLSI design and CAD education in
Taiwan. His current research focus is on design technology for System-on-a-Chips (SOC)
targeted towards multimedia applications.
He has served on program committee, organizing committee, steering committee, and
executive committee for several conferences and workshops on various aspects of CAD
including the Design Automation Conference(DAC), the International Conference on
CAD(ICCAD), the Asia South-Pacific Conference on Design Automation(ASP-DAC), the
Design, Automation, & Test of Europe (DATE) Conference, the Asia Pacific Conference on
Hardware Description Languages (APCHDL), the workshop on Synthesis And System
Integration of MIxed technology(SASIMI), the International Symposium on System
9
Synthesis(ISSS), the International Symposium on Physical Design (ISPD), the ACM
Embedded Software Conference (EMSOFT), the International Symposium on VLSI Design,
Automation and Test (VLSI-DAT), the International SOC Conference (ISOCC), and the
VLSI/CAD Symposium of Taiwan. Professor Lin has also served on the editorial boards of the
ACM Transactions on Design Automation of Electronic Systems (TODAES) , the ACM
Transactions on Embedded Computing Systems (TECS) , Journal of the Chinese Engineering
Society, and the Journal of Information Science and Engineering (JISE) .
Dr. Lin has been a consultant to the Institute for Information Industry (III), the Industrial
Technology Research Institute (ITRI), the Electronic Research and Service Organization
(ERSO), and the Computer and Communication research Laboratory (CCL). He is an advisor
of the Science and Technology Advisory Office, Ministry of Education, R.O.C., in charge of
VLSI design education program, and a co-PI of the Chip and System Implementation Center
(CIC) program of the National Science Council (NSC), R.O.C. He also served on Intel's Asia-
Pacific Advisory Board and advisory boards of several EDA and consumer electronics startups.
He is an independent board member of Etron Technology Inc.
Professor Lin received a national fellowship for studying abroad from the Ministry of Education,
R.O.C. in 1983, co-received the Outstanding Young Author Award from the IEEE Circuit and
System Society in 1990, the Outstanding Design Award from the ASP-DAC University LSI
Design Contest in 2000, and received the Highest-Honored Research Award from the NSC
three consecutive times in 1992, 1994, and 1996. He has been an NSC research fellow since
1998. In 2007, he received the first Industrial Contribution Award from the Ministry of Economic
Affairs, R.O.C., for his impact on Taiwan IC design industry.
Dr. Lin is a senior member of the IEEE, the IEEE Computer Society, the IEEE Circuit and
System Society, the IEEE Solid State Circuit Society, and the Association for Computing
Machinery. He is a board member of the Taiwan IC Design Society. He is elected the president
of Taiwan Integrated Circuit Design Society (TICD) for 2005-06. He is a nation-certified
electrical engineer.
Abstract
AI Innovation Research Program of Taiwan
Starting from 2017, Taiwan’s Ministry of Science and Technology (MOST) initiates a 5-year
USD500M program for AI innovation research. Its goals include: incubating high-quality
talents, developing cutting-edge technology, and empowering industry with AI technology.
Taiwan has been known for strong at electronics hardware manufacturing but weak at
software. The on-going AI revolution is changing the existing paradigm. We seek international
collaborations that complement each other. In this talk, I will briefly introduce the program
contents and ask for your comment/advice.
10
Tsung-Yi Ho
Department of Computer Science, National Tsing Hua University
e-mail: [email protected]
phone number: +886-922-271-858
Tsung-Yi Ho received his Ph.D. in Electrical Engineering from National Taiwan University in
2005. He is a Professor with the Department of Computer Science of National Tsing Hua
University, Hsinchu, Taiwan. His research interests include design automation and test for
microfluidic biochips and nanometer integrated circuits. He has presented 10 tutorials and
contributed 10 special sessions in ACM/IEEE conferences, all in design automation for
microfluidic biochips. He has been the recipient of the Invitational Fellowship of the Japan
Society for the Promotion of Science (JSPS), the Humboldt Research Fellowship by the
Alexander von Humboldt Foundation, and the Hans Fischer Fellow by the Institute of Advanced
Study of the Technical University of Munich. He was a recipient of the Best Paper Awards at
the VLSI Test Symposium (VTS) in 2013 and IEEE Transactions on Computer-Aided Design
of Integrated Circuits and Systems in 2015. He served as a Distinguished Visitor of the IEEE
Computer Society for 2013-2015, the Chair of the IEEE Computer Society Tainan Chapter for
2013-2015, and the Chair of the ACM SIGDA Taiwan Chapter for 2014-2015. Currently he
serves as an ACM Distinguished Speaker, a Distinguished Lecturer of the IEEE Circuits and
Systems Society, and Associate Editor of the ACM Journal on Emerging Technologies in
Computing Systems, ACM Transactions on Design Automation of Electronic Systems, ACM
Transactions on Embedded Computing Systems, IEEE Transactions on Computer-Aided
Design of Integrated Circuits and Systems, and IEEE Transactions on Very Large Scale
Integration Systems, Guest Editor of IEEE Design & Test of Computers, and the Technical
Program Committees of major conferences, including DAC, ICCAD, DATE, ASP-DAC, ISPD,
ICCD, etc.
11
Dominik Ślęzak
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw
e-mail: [email protected]
phone number: +48 660 495 076
Dominik Ślęzak received his Ph.D. in Computer Science in 2002 from University of Warsaw,
Poland, and D.Sc. in 2011 from Institute of Computer Science of Polish Academy of Sciences.
After finishing his Ph.D. he started working as Assistant Professor in Polish-Japanese
Academy of Information Technology, Poland, and then at University of Regina, Canada.
In 2005 he suspended his academic career and co-founded Infobright, where he worked as
Chief Scientist until 2017. In the meantime, in 2009 he re-joined University of Warsaw, where
he currently holds full-time position of Associate Professor in Institute of Informatics. He also
cooperated as Adjunct Professor with McMaster and York University in Canada.
Dominik is continually involved in both academic and industry R&D projects, in the areas of
Artificial Intelligence, Machine Learning and Knowledge Discovery. His research has strong
roots in Rough Set Theory co-developed by his scientific supervisor and mentor, Professor
Andrzej Skowron. His interests have been always related to Data Exploration and Decision
Support tools and methods that would be easy to use by domain experts and practitioners.
In Infobright he co-designed Analytical Database Engine aimed at fast low-resource
processing of massive volumes of machine-generated data, based on the paradigms of
Columnar Data Stores and Rough Set Approximations. In 2015 he co-established the team
working on highly scalable Approximate Analytics Engine based on the elements of
Approximate Computing and Information Granulation, whose development is now continued
in cooperation with Security on Demand. In parallel, since 2012 he has been active as
co-organizer of open data mining challenges at Knowledge Pit. He also played leading roles
in several government-funded projects including SYNAT and DISESOR.
Dominik co-edited over 20 books and volumes of conference proceedings. He co-authored
over 150 papers for books, journals and conferences. He is co-inventor in five US patents.
He delivered invited plenary talks at over 20 international scientific events. He also serves
actively for the community. He is Associate Editor / Editorial Board Member for several journals
including Information Sciences and Intelligent Information Systems (both since 2007). He was
one of Founding Editors of Communications in Computer and Information Science.
He co-organized over 20 international conferences including a series of FedCSIS/AAIA
Symposia on AI Applications (2013-2017). In 2012-2014 he served as President of
International Rough Set Society (IRSS). In 2015 he was appointed as IRSS Fellow. In 2014
he received Outstanding Service Award from Web Intelligence Consortium. Since 2014 he has
served in IEEE Technical Committee on Intelligent Informatics.
12
Abstract
Toward Approximate Intelligence – Lessons Learnt from Development of KDD Methods
and Approximate Query Engines
Artificial Intelligence methods are currently regaining a lot of attention in the areas of data
analytics and decision support. Given the increasing amount of information and computational
resources available, it is now possible for intelligent algorithms to learn from the data and assist
humans more efficiently. Still, there is a question about the goals of learning and a form of the
resulting data-driven knowledge. It is evident that humans do not operate with fully exact /
precise information in decision-making and, therefore, it might be unnecessary (or even
inadvisable) to provide them with complete outcomes of analytical processes. Consequently,
the next question arises whether incomplete (yet sufficiently rich) results of computations on
large data could be delivered faster than their standard counterparts, in order to improve
interactions between humans and intelligent processes. Such questions – referring to human-
computer interactions – are analogous to questions about precision of calculations conducted
internally by machine learning and knowledge discovery methods, whereby various heuristic
algorithms could be accelerated by letting them rely on approximate computations. This leads
us toward discussion on the importance of approximations in the areas of machine intelligence
and business intelligence and, more broadly, the meaning of approximate representations and
computations for various aspects of Artificial Intelligence. In this talk, we will refer to this
discussion using two cases studies – an example of approximate database software developed
using the paradigms of rough-granular computing and an example of heuristic feature selection
method that was successfully applied in several KDD projects.
13
Chang-Shing Lee
Department of Computer Science and Information Engineering, National University of Tainan
e-mail: [email protected]
phone number: +886-6-2602573
Chang-Shing Lee (SM’09) received the Ph.D. degree in Computer Science and Information
Engineering from the National Cheng Kung University, Tainan, Taiwan, in 1998. He is currently
a Professor with the Department of Computer Science and Information Engineering, National
University of Tainan and he was the Dean of Research and Development Office from January
2011 to July 2015. His current research interests include adaptive assessment, intelligent
agent, ontology applications, Capability Maturity Model Integration (CMMI), fuzzy theory and
applications, and machine learning. He also holds several patents on Fuzzy Markup Language
(FML), ontology engineering, document classification, image filtering, and healthcare. He was
awarded Certificate of Appreciation for outstanding contributions to the development of IEEE
Standard 1855TM-2016 (IEEE Standard for Fuzzy Markup Language). He was the Emergent
Technologies Technical Committee (ETTC) Chair of the IEEE Computational Intelligence
Society (CIS) from 2009 to 2010 and the ETTC Vice-Chair of the IEEE CIS in 2008. He is also
an Associate Editor or Editor Board Member of International Journals, such as IEEE
Transactions on Computational Intelligence and AI in Games (IEEE TCIAIG), Applied
Intelligence, Soft Computing, Journal of Ambient Intelligence & Humanized Computing (AIHC),
International Journal of Fuzzy Systems (IJFS), Journal of Information Science and Engineering
(JISE), and Journal of Advanced Computational Intelligence and Intelligent Informatics
(JACIII). He also guest edited IEEE TCIAIG, Applied Intelligence, Journal of Internet
Technology (JIT), and IJFS. Prof. Lee was awarded the outstanding achievement in
Information and Computer Education & Taiwan Academic Network (TANet) by Ministry of
Education of Taiwan in 2009 and the excellent or good researcher by National University of
Tainan from 2010 to 2016. Additionally, he also served the general co-chair of 2015 IEEE
Conference on Computational Intelligence and Games (IEEE CIG 2015), the general chair of
the 2015 Conference on Technologies and Applications of Artificial Intelligence (TAAI 2015),
the program chair of the 2011 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE
2011), and the competition chair of the FUZZ-IEEE 2013, the competition co-chair of the FUZZ-
IEEE 2015, FUZZ-IEEE 2017, and 2016 IEEE World Congress on Computational Intelligence
(IEEE WCCI 2016) and IEEE WCCI 2018. He is also a member of the Program Committees
of more than 50 conferences. He is a senior member of the IEEE CIS, a member of the
Taiwanese Association for Artificial Intelligence (TAAI), and the Software Engineering
Association Taiwan. He was a member of the standing committee of TAAI from 2011 to 2016
and one of the standing supervisors of Academia-Industry Consortium for Southern Taiwan
Science Park from 2012 to 2013.
14
Abstract
Intelligent Agent for Human and Machine Co-Learning on Game of Go
In this talk, we will demonstrate the application of Fuzzy Markup Language (FML) to construct
an FML-based Dynamic Assessment Agent (FDAA), and we present an FML-based Human–
Machine Cooperative System (FHMCS) for the game of Go. The proposed FDAA comprises
an intelligent decision-making and learning mechanism, an intelligent game bot, a proximal
development agent, and an intelligent agent. The intelligent game bot is based on the open-
source code of Facebook’s Darkforest, and it features a representational state transfer
application programming interface mechanism. The proximal development agent contains
a dynamic assessment mechanism, a GoSocket mechanism, and an FML engine with a fuzzy
knowledge base and rule base. The intelligent agent contains a GoSocket engine and
a summarization agent that is based on the estimated win rate, real-time simulation number,
and matching degree of predicted moves. Additionally, the FML for player performance
evaluation and linguistic descriptions for game results commentary are presented.
We experimentally verify and validate the performance of the FDAA and variants of the FHMCS
by testing five games in 2016 and 60 games of Google’s Master Go, a new version of the
AlphaGo program, in January 2017. The experimental results demonstrate that the proposed
FDAA can work effectively for Go applications.
15
Krzysztof Krawiec
Faculty of Computing, Poznan University of Technology
e-mail: [email protected]
phone number: +48 61 66 53 061
Krzysztof Krawiec. Associate professor in the Institute of Computing Science, Poznan
University of Technology. Past work includes program synthesis, in particular using genetic
programming; deep learning for image analysis and game playing; evolutionary computation
for machine learning, games and pattern recognition; coevolutionary algorithms and test-based
problems. K. Krawiec is an author of over 100 papers on the above topics, an associate editor
of the Genetic Programming and Evolvable Machines journal, and formerly a visiting
researcher at University of California and Massachusetts Institute of Technology. More at
http://www.cs.put.poznan.pl/kkrawiec
Abstract
Deep Feature Engineering for Segmentation, Anomaly Detection
and Diagnosing in Medical Imaging
Deep neural networks facilitate composing effective models from a rich repertoire of
components: feature maps, embeddings, convolutions, etc., allowing so to address a range of
tasks of practical interest. In my talk, I will illustrate this characteristic of contemporary neural
models using three image analysis problems we approached in my group: segmentation of the
blood vessel network in ophthalmology (for 2D fundus imaging and 3D optical coherence
tomography), anomaly detection in computer tomography, and diagnostic decision support for
fluorescein angiography. Problems, methods, neural architectures, and main results will be
presented, followed by a brief discussion of future outlooks and related ongoing projects.
16
Hsun-Ping Hsieh
Department of Electrical Engineering, National Cheng Kung University
e-mail: [email protected]
phone number: +886-6-2757575 ext. 62322
Hsun-Ping Hsieh is now an Assistant Professor and leads Urban Science and Computing Lab
(UCLAB) at Department of Electrical Engineering, National Cheng Kung University, Taiwan.
He received his Ph.D. degree in Graduate Institute of Networking and Multimedia at National
Taiwan University. Hsun-Ping’s international recognition includes ACM KDD Cup 2010 First
Prize, Garmin Fellowship 2014 and 2013, and Best Intern Award at Microsoft Research Asia.
His research interests include big data mining, urban computing, and AI-based smart city
services. In the past years, he published a series of papers in location-based services and
urban computing, including ACM TIST, KAIS, ACM WWW, ACM SIGKDD, ACM SIGIR, ACM
MM, IEEE ICDM, PKDD, AAAI ICWSM and ACM CIKM. He had been the main tutorial
presenter on the topic of social media analytics at PAKDD 2016, WWW 2015, and on route
planning at ICWSM 2014 and ASONAM 2014. Hsun-Ping’s research is interdisciplinary from
the perspective of Data Science. In 2016, he won EIGHT best dissertation awards from several
leading academia associations and industrial companies in Taiwan.
Abstract
Mining Geo-Social Media and Urban Informatics
for Smart Environmental and Business Applications
With the maturity of location-acquisition techniques and mobile devises, a novel type of online
services, geo-social media (e.g. Foursquare, Facebook, and Flickr), have caught not only
users’ attention but also researchers' efforts on solving real-world problems. People are
allowed to keep track of what they see, where they go, who they know, what they experience,
and when they do anything in the forms of images, locations, social network, short texts, and
time stamps respectively. On the other hand, we deploy sensors such as weather stations,
invigilators or vehicle detector to collect urban informatics in the physical environment of the
city. In this talk, we seek to point out and discuss a new research direction that connects
geo-social media, urban sensor data to enable novel environmental and business applications.
We develop a general-purpose spatial-temporal inference framework to not only model
user-environment interactions, but also is preliminarily validated to be capable of accurately
infer air quality, noise degree, and commercial foot traffic. Such promising results are believed
to encourage a brave new direction for making a city smart fusing geo-social media and urban
informatics.
17
Jakub Nalepa
Department of Automatic Control, Electronics and Computer Science, Silesian University of
Technology
Future Processing, Gliwice, Poland
e-mail: [email protected], [email protected]
phone number: +48 32 237 2117
Jakub Nalepa received a M.Sc. degree (2011) and a Ph.D. degree (2016), both with distinction
in computer science from the Silesian University of Technology, Gliwice, Poland. Dr. Nalepa
is currently a researcher at the Department of Automatic Control, Electronics and Computer
Science, Silesian University of Technology. Since 2010, he has been a research scientist at
Future Processing, Gliwice, Poland, where he is currently the lead research scientist in
machine learning and medical imaging projects. Dr. Nalepa's research interests encompass
machine learning, evolutionary algorithms, especially (adaptive and self-adaptive) genetic and
memetic algorithms, pattern recognition, optimization, medical imaging, parallel computing,
deep learning and interdisciplinary applications of these methods. He has been involved in
several projects related to the above-mentioned domains – in both academia and industry.
So far, Dr. Nalepa published more than 60 papers in these fields. Dr. Nalepa received the best-
paper awards at the 3PGCIC 2016 and IBERAMIA 2016 conferences, and was nominated to
the best-paper award at the EvoCOP 2017 conference. Dr. Nalepa, together with Michael
P. Hayball (CEO of Cambridge Computed Imaging Ltd), gave an invited talk on applying deep
networks in medical imaging at the University of Cambridge (at the Centre for Mathematical
Imaging in Healthcare seminar, 2017). He is a reviewer for several international journals and
conferences. Dr. Nalepa is an IEEE and ACM member.
Abstract
Towards Automatic Design of Deep Neural Networks Through Evolutionary Computation
Introducing deep neural networks (DNNs) was undoubtedly a breakthrough in artificial
intelligence. DNNs have achieved remarkable success in a plethora of tasks. However, the
performance of such deep systems strongly depends on their hyper-parameters and
architectures which often must be selected by an experienced practitioner. Selection of
hyper-parameters remains a substantial obstacle in designing DNNs in practice and it is a very
time-consuming process. In this talk, we show how these hyper-parameters can be evolved
using a particle swarm optimization (PSO) technique [1]. We demonstrate that PSO efficiently
explores the space of hyper-parameter values, allowing DNNs of a minimal topology to obtain
competitive classification performance over the multi-class MNIST dataset. We showed that
very small DNNs optimized by PSO retrieve promising classification accuracy for a challenging
CIFAR-10 dataset. Importantly, PSO can be successfully used to improve the performance of
existent deep topologies. Extensive experimental study, backed-up with the statistical tests,
18
revealed that PSO is an effective technique for automating hyper-parameter selection and it
efficiently exploits computational resources.
References
1. Lorenzo, P.R., Nalepa, J., Kawulok, M., Ramos, L.S., Pastor, J.R.: Particle swarm
optimization for hyper-parameter selection in deep neural networks. In: Proceedings of
the Genetic and Evolutionary Computation Conference. pp. 481-488. GECCO'17,
ACM, New York, NY, USA (2017), http://doi.acm.org/10.1145/3071178.3071208
19
Jason S. Chang
Department of Computer Science, National Tsing Hua University
e-mail: [email protected]
phone number: +886-3-5731069
Prof. Jason S. Chang has served as Professor of Computational Linguistics at National Tsing
Hua University, Taiwan. He has working on better ways of using technology to enhance second
language learning. Under the CANDLE projects, significant progress was made on developing
novel online tools, including a fast trillion-word linguistic search engine Linggle, and an
interactive writing environment WriteAhead based on induced grammars.
Prof. Chang was a founding member and President of Association for CLCLP and member of
Standing Committee on Language Policy, Ministry of Education (MOE). He has served as
Program Committee Area Chair, the Annual Conference of the Association for Computational
Linguistics (ACL), and on PC of numerous international conferences, as well as Advisory Board
of the Journal of CLCLP.
Prof. Chang earned a B.S. in Mechanical Engineering and an M.S. in Computer Science from
NTHU. He then went to New York University and obtained Ph.D. (1986) in Computer Science.
His publications include over a hundred referred papers in international conferences, technical
journals, including the prestigious Computational Linguistics published by MIT Press. When
not busy researching and writing code, he writes monthly columns in Scientific American,
Taiwanese Editions.
Abstract
Inducing Lexical Grammar from Monolingual and Bilingual Corpora
for Computer Assisted Language Learning and Translation
The past decade has witnessed the emergence of corpus-based statistical methods for
automatic error correction and essay rating with ever increasing performance. However, the
goal of fully automatic, high-quality Grammar Error Correction (GEC) is still elusive. In this talk,
I will describe our ongoing work on induction of lexical Pattern Grammar (PG) as well as
Synchronous Pattern grammar (SPG) for tasks such as Interactive Writing Environment,
Grammatical Error Correction, Computational Lexicography, and Machine Translation.
As example applications of this line of research, I will demonstrate a linguistic search engine
Linggle and an interactive writing system, WriteAhead, aimed at helping language learners by
providing writing help as solicited or unsolicited search results. Linggle accepts queries with
keywords, wildcard, wild part of speech (PoS), synonymous words, and additional regular
expression (RE) operators, and returns bundles with frequency counts. WriteAhead provides
L2 learners with writing and editing prompts to help them write fluently and accurately. For that,
we automatically analyze a given corpus, and extract grammar patterns and common error-
and-edit patterns. At run-time, as learners type (or mouse over) a word, the system
automatically retrieves and displays grammar patterns and examples, most relevant to the
word. The user can opt for patterns from a general corpus, academic corpus, learner corpus,
20
or common error corpus. WriteAhead proactively engages the user by providing steady, timely,
and relevant grammar patterns (or synchronous grammar patterns) for effective assisted
writing and translation.
21
Grzegorz J. Nalepa
Department of Applied Computer Science, AGH University of Science and Technology;
Jagiellonian University
e-mail: [email protected]
phone number: +48 12 61 73 856
Grzegorz J. Nalepa (http://gjn.re) is an engineer with degrees in computer science – artificial
intelligence, and philosophy. He has been working in the area of intelligent systems and
knowledge engineering for over 15 years. He formulated the eXtended Tabular Trees rule
representation method, as well as the Semantic Knowledge Engineering approach.
He authored a book "Modeling with Rules using Semantic Knowledge Engineering" (Springer
2017). He co-edited a book "Synergies Between Knowledge Engineering and Software
Engineering" (Springer 2017). He co-authored over 150 research papers in international
journals and conferences. He coordinates GEIST - Group for Engineering of Intelligent
Systems and Technologies (http://geist.re) cooperating with AGH University and Jagiellonian
University in Krakow, Poland. For almost 10 years he has been co-chairing the Knowledge and
Software Engineering Workshop (KESE) at KI, the German AI conference, Spanish CAEPIA,
as well ECAI. He is the President of the Polish Artificial Intelligence Society (PSSI), member
of EurAI. He is also a member of IEEE, Italian Artificial Intelligence Society (AI*IA), KES, Polish
Cognitive Science Society (PTK). His recent interests include context-aware systems and
affective computing.
Abstract
Context, Autonomy, Affect. Methodological and Ethical Challenges of Today's AI
Research in the area of pervasive computing and ambient intelligence aims to make use of
context information to allow devices or applications behave in a context-aware, thus
“intelligent” way. By a classic definition context is any information that can be used to
characterize the situation of an entity. Cognitive assistants, or more recently smart advisors,
are computer systems that augment human capabilities by providing personalized context-
driven decision support. Today, such systems use wearable devices and ambient intelligence
technologies to be constantly available to the user. Affective Computing is a field of study that
puts interest in the design and description of systems that are able to collect, interpret, and
process emotional states. Assuming that emotions are physical and cognitive and as such they
can be studied interdisciplinary by computer science, bio-medical engineering, cognitive
science, and psychology. Research challenges in this area include detection and classification
of affective states, as well as inducement of emotions in laboratory studies. We argue, that
human-centric cognitive computer systems should be able to detect and interpret changes of
emotional state of the users. To this goal we propose to combine to computing approaches:
22
affective computing and context-aware systems. Context-awareness boosts opportunities for
cognitive assistants to be autonomous, whereas emotion-awareness can make them more
human-like in the interaction with the users. The practical development of such systems poses
number of methodological challenges, including the management of big data, bridging of
symbolic knowledge-based approaches with machine learning methods, as well as providing
explanation capabilities. Finally, the design of these technologies raises important ethical
concerns and societal issues. Such AI systems should be understandable, transparent, and
safe, to have positive economical and societal impact.
23
Ming-Feng Tsai
Department of Computer Science, National Chengchi University
e-mail: [email protected]
phone number: +886-2-2939-3091 ext. 62558
Ming-Feng Tsai is currently an Associate Professor in the Department of Computer Science at
National Chengchi University. In 2006, he was at Microsoft Research Asia as an intern with
the Web Search & Mining Group, and was awarded by the research institution the Best Intern
of the Year, and invited to visit the headquarters of Microsoft and Bill Gates’ house in
Redmond. He received his Ph.D. degree from National Taiwan University in 2009. After
receiving his Ph.D. degree, he worked at National University of Singapore as a Research
Fellow, participating in a research project related to machine translation. In 2010, sponsored
by Ministry of Scienceand Technology, he joined University of Illinois at Urbana-Champaign
as a visiting scientist, working on the projects associated with advanced Web search and
mining. His research interests span the areas of information retrieval, machine learning,
recommender systems, natural language processing, computational social science.
Abstract
Textual Data Analytics in Finance
The growing amount of public financial data makes it more and more important to learn how
to discover valuable information for financial decision-making. This talk presents our recent
studies on exploring and mining soft information in finance, which usually refers to text,
including opinions, ideas, and market commentary. This talk will cover several machine
learning techniques, such as learning to rank and word embedding, on financial reports for the
study of financial risk among companies, and the demonstrations on our newly developed
web-based information system, Fin10K, to showcase its ability for textual analytics in finance.
24
Radosław Gwarek
11 bit studios SA
e-mail:[email protected]
phone number: +48 504 721 801
Working in the games industry since 2011. Designer of AI systems for a number of successful
games, including This War of Mine and Frostpunk.Lecturer at the Warsaw Film School,
teaching new generations of video game developers creating games using Unity Engine,
basics of game programming and balancing combat systems for games.
Deadly Monsters (2017)
A modification for PC version of Minecraft that adds new critters, featuring new and unique AI
behaviours of enemies, which require a new approach to fighting them. The entire modification
was created using Java programming language.
Pixwing (2015)
Mobile game in which the player pilots an airplane by moving their phone or tablet. The game
tracks the movements using accelerometers, gravity sensors, gyroscopes and rotational vector
sensors. Virtual Reality helmets utilise the same motion tracking technology, which allowed
me to use sensor fusion algorithms from VR helmets.
Frostpunk (2015-2017)
City building game in post-apocalyptic setting, including elements of society simulation. Design
of complex AI system for multiple agents, design of necessary tools for AI development (editor,
debugger). Creating behaviours according to the requirements of the story designers.
Implementing my own ideas for additional behaviours. Maintaining test cases in order to keep
the game error-free and stable.
This War of Mine, This War of Mine: The Little Ones (2013-2015)
Survival game about the fate of civilians in a modern conflict. More traditional approach to
gameplay AI with a smaller number of agents. Design of AI editor and debugging tools.
Flix (2013)
A logic game for Android systems. Players use puzzles to create a path for their character.
The main goal is to reach the finish line, while avoiding obstacles and enemies. Programming,
graphics, overall design.
25
Anomaly 2 (2012)
Reversed tower defense game, in which player controls the attackers. Technical side of the
design, creating scripts supporting additional events during the game such as tutorial or
interactive cutscenes. Diagnostics and/or fixing of more important problems reported by
Quality Assurance Department.
Funky Smugglers (2011)
Casual arcade game for mobile platforms and PCs. General design of the game, particularly
the balancing of the economy values and the difficulty levels.
Abstract
Meta AI – The artificial mind
Complex AI systems used in video games can exhibit unexpected behaviours. This is usually
treated as an error, but in some circumstances the results are surprisingly good, leading to AIs
behaving in their own way, performing lifelike actions that have never been designed.
Examples will be based on the game “This War of Mine”, where we will look deeper into the
system and see how characters make their decisions. Quickly switching between multiple parts
of the behaviour tree modules may produce unique behaviour that exists only in a specific time
frame. During this state, the agent can perform additional actions that, although resulting from
error, can be perceived as humanlike behaviour. The whole process can be compared to
evolution, with the developer playing the role of natural selection, eliminating critical errors and
keeping the “good” ones – quite similar to how harmful mutations are weeded out of the gene
pool and beneficial ones propagate. The final effect is a new, artificially created set of
behaviours which are a result of random errors.
26
Piotr Babieno
CEO Bloober Team SA
+48 662 216 616
Mr. Babieno has overall over a decade of professional leadership experience in the field of
gaming industry, business development and sales. He studied at Technical University of
Krakow, Warsaw School of Film and Polish Open University in Kraków. Since 2006, he has
served as managing developer in companies creating games on next generation consoles.
Mr. Babieno is a lecturer, and the co-creator of the program and the syllabus of the
postgraduate course “European Academy of Games” organized by Jagiellonian University in
collaboration with the Academy of Mining and Metallurgy. He is also a member of the Program
Board of EAG. At the invitation of the Government of Canada he was a lecturer at Jalloo
Festival of Animation and Gaming. In 2009 thanks to extensive experience and knowledge of
new technologies, Mr. Babieno was appointed an expert for “Technology Perspectives Krakow
– Malopolska 2020” – Foresight.
Abstract
Hidden Horror and Catharsis 2.0
Most horrors tell a story, hidden horror tackles certain subjects, most often philosophical
or psychological in their nature to provide an experience. Each part of the game, from the
plot to level design, works to strengthen and envision the subject.
Horror by definition needs to provide a feeling of relief at its end; HH goes one step further
and provides a cathartic experience that goes beyond just an emotional relief – the
dilemmas and decisions presented in the game should make you ponder on what kind of
person you are. We always have a point of view on each and every aspect of our lives
– this tells us a lot about ourselves.
The dilemmas we design for our games are not abstract, they live in our imagination as
something we know exists but happens to the friends of our friends, or we read about them
in newspapers. Living through those experiences in a game may prepare us to face such
situations in real life – think of them as preventive coaching or even therapy.
Catharsis (1.0) is the process of releasing, and thereby providing relief from, strong or
repressed emotions. Most horror games revolve around this concept – giving plenty of fear
to people who are willing to live it through and feel stronger because of it. We went deeper
than that – we still provide tons of scares for people, but what we also do is we add a lot of
moral dilemmas during that frightening journey. The catharsis 2.0 is not only about being
relieved from repressed feelings – it’s a journey to understand not only emotions but also
our way of reasoning – how we see the world and what makes us tick. A kind of introvert
relief – we have a lot of bottled layers of our psyche that we do not explore, they may lay
27
dormant even our entire lives – with Catharsis 2.0 we want to provide those experiences
to people.
Horror has always been most popular during times of biggest social tensions. It’s role was
always cathartic, so it already had a deeper cultural layer to it in comparison to other
genres. Horror is also meant for older audiences, which obviously is part of our main target,
because of the level of their understanding of the world. And the baggage of experiences
that burdens their shoulders.
Hidden Horror is considered as a therapeutic tool by a friendly scientific unit, and one of
the players is writing a doctoral dissertation on overcoming his depression because of his
experience with Layers of Fear.
28
Sheng-Jyh Wang
Department of Electronic Engineering, National Chiao Tung University
e-mail: [email protected]
phone number: +886-3-5712121 ext. 54177
Sheng-Jyh Wang (M’95) received the B.S. degree in electronics engineering from National
Chiao Tung University, Taiwan, in 1984, and the M.S. and Ph.D. degrees in electrical
engineering from Stanford University, USA, in 1989 and 1994, respectively.
He is currently a professor with the Institute of Electronics, NCTU. His research interests are
image processing, video processing, image analysis, and machine learning.
Abstract
An Efficient Sequential Partitioning Method for Clustering and Manifold Learning
How to automatically extract useful information from data has long been a crucial issue in
plentiful scientific studies, especially in the era of big data. In this presentation, we will mention
how we modify over the Bayesian Sequential Partitioning (BSP) algorithm to develop a new
algorithm that can be used for data clustering and manifold learning. The BSP algorithm was
originally designed for the estimation of probability density function in the feature space,
especially in a high-dimensional case. By properly modifying the score function in the BSP
algorithm, we turn the original BSP algorithm into an effective method for data grouping and
for the exploration of data manifolds. We also develop a hierarchical framework to analyze the
correlations among partitioned data groups at different scales. By applying the multi-scale data
analysis to image segmentation, we develop a hierarchical image segmentation algorithm that
outperforms the state-of-the-arts image segmentation algorithms.
29
Cezary Dołęga
Neurosoft Sp. z o.o.
e-mail: [email protected]
phone number: +48 601 774 737
Co-founder and co-owner of Neurosoft Sp. z o.o. Master of Electronics Engineering, graduated
from Wroclaw University of Technology in 1990. Researcher at the Institute of Technical
Cybernetics from 1990 to 1999. Since the beginning of Neurosoft – Vice-President, Chief
Technology Officer. Served as Scientific Project Manager of numerous research projects,
including:
ISKIP – Intelligent Comprehensive System Vehicle Identification. PL: Operational Program
Innovative Economy, UDA-POIG.01.03.01-00-147/11 (2011-2013);
ISM3P – Intelligent monitoring and enforcement system for overloaded vehicles.
PL: National Centre for Research and Development, Applied Research Programme,
no. PBS1/B3/4/2012 (2012-2014);
MOBIS – Methods of pedestrian safety assessment using video analysis. PL National
Centre for Research and Development, Applied Research Programme,
no. PBS1/A2/8/2012 (2013-2015);
OptiCITIES – Optimise Citizen Mobility and Freight Management in Urban Environments.
EU: 7th Framework Program, FP7-SST-2013-RTD-1 (2013-2016);
AUDIOSCOPE – System for automatic keyword spotting in spontaneous speech.
PL: National Centre for Research and Development, Applied Research Programme,
Applied Research Programme, no. PBS3/A9/36/2015 (2015-2017);
NEUROPARK – Intelligent Multi-Sensor Parking Lot Management System.
PL: National Centre for Research and Development, no. POIR.01.01.01-00-0472/15
(2015-2017);
MOVIESTA – Mobile Self-Adjusting Video-based Detections System for Traffic
Applications. EU: Eurostars Programme, Project id: 10196 (2016-2018);
NEUROSPACE – Intelligent Modeling and Monitoring System for Infrastructure Spatial
Systems, PL: National Centre for Research and Development, no. POIR.01.01.01-00-
1176/15-00 (2016-2018);
NEUROWIM – Intelligent Vehicle Weigh-in-Motion System with Enhanced Accuracy.
PL: National Centre for Research and Development, no. POIR.01.01.01-00-0612/16
(2017-2018).
30
Abstract
NeuroCar – Applications of artificial intelligence in Intelligent Transport Systems:
traffic safety, intelligent infrastructure and autonomous vehicles
The future of transport systems will certainly be closely linked to the development of "smart"
technologies that will be used in various areas. First, it is necessary to make more effective
use of existing infrastructure (road network, parking, urban space) by optimizing the behavior
of its users. Secondly, a constant trend is to increase transport comfort, e.g. by raising safety
levels or automating processes (like driver-less vehicles). Thirdly, transport activity begins to
function in new, previously unused spaces such as urban airspace (drones) or outer space.
In each of these cases, artificial intelligence methods are more and more widespread, and in
particular artificial neural networks.
Neurosoft has been developing the NeuroCar product line for over 10 years, in which artificial
intelligence algorithms have been practically applied. In the beginning the NeuroCar VI
(Vehicle Identification) system will be presented. The system is used to automatically detect
and identify vehicles (vehicle detection, vehicle class identification, license plate recognition,
country of origin recognition, make and model recognition, determination of vehicle movement
parameters such direction and speed, detection of dangerous goods). The applications of this
type of system for traffic measurement and road safety improvement will be presented.
The second will be the NeuroCar RL (Red Light Enforcement) system, which is used to detect
red light traffic violations. The third will be the NeuroCar PARK (Parking Lot Management)
system that is used to determine the occupancy of parking spaces, detection of free parking
spots and the parking lots occupancy optimization. The NeuroCar Flow for traffic
measurement, which uses semantic image segmentation and object tracking algorithms (deep
neural networks), will be the latest to be presented.
The last part of the presentation will contain brief overview of new research projects at
Neurosoft: NeuroSpace – an autonomous multi-rotor UAV for road infrastructure monitoring;
NeuroWIM – high accuracy vehicle Weigh-in-Motion system and MoVieSta – an intelligent road
infrastructure to support self-steering vehicles.
31
Winston Hsu
Department of Computer Science and Information Engineering, National Taiwan University
e-mail: [email protected]
phone number: +886-2-3366-4888 ext. 512
Prof. Winston Hsu is an active researcher dedicated to large-scale image/video
retrieval/mining, visual recognition, and machine intelligence. He is keen to realizing advanced
researches towards business deliverables via academia-industry collaborations. He is
a Professor in the Department of Computer Science and Information Engineering, National
Taiwan University. He founded MiRA (Multimedia indexing, Retrieval, and Analysis) research
group and co-leads Communication and Multimedia Lab (CMLab). Working closely with the
industry, he was a Visiting Scientist at Microsoft Research Redmond (2014) and had his
sabbatical leave (2016-2017) at IBM TJ Watson Research Center, New York, to enhance
Watson’s visual cognition, where he contributed the first AI produced movie trailer. He is the
founding Director for NVIDIA AI Lab (NTU), the 1st in Asia. He received Ph.D. (2007) from
Columbia University, New York. Before that, he was a founding engineer and research
manager in CyberLink Corp, now a public image/video software company. He serves as the
Associate Editors for IEEE Multimedia Magazine and IEEE Transactions on Multimedia, two
premier journals.
Abstract
Learning from Multimodal Data Streams
Images, videos, audios, and user-contributed logs, etc., are major data types nowadays
essential for disruptive opportunities in social media, entertainments, education, healthcare,
IoT, etc. However, the developed techniques are far behind the dire needs. For leveraging
multimodal data streams with effective deep neural networks, we will present advanced and
novel methods for jointly considering spatial and sequential neural networks and their
variations, which well approximate the multimodal data streams. We will show the importance
and the problems for cross-domain learning – computing data of different types in the same
semantic space – and present several solutions. We will demonstrate how to utilize different
modalities for improving challenging learning tasks in the end-to-end neural networks. We will
showcase some recent works published in the top venues e.g., CVPR, ICCV, etc.
32
Bartosz Wilczyński
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw
e-mail: [email protected]
phone number: +48 22 55 44 577
Bartosz Wilczyński is an assistant professor in the Institute of Informatics since 2011.
He obtained his PhD in Mathematics from Polish Academy of Science and has worked also in
Lawrence Livermore National Laboratory in USA and European Molecular Biology Laboratory
in Heidelberg.
His research is focused on applying computational methods to the problems arising in modern
genomics. He worked on mathemetical models of regulatory networks, Bayesian Networks and
applications of machine learning methods to functional genomic element identification. He is
the principal investigator in a number of projects funded by grants including EMBO Installation
Grant, ERA-NET European grant, and multiple national projects by Polish National Science
Centre and Foundation for Polish Science.
Abstract
Machine learning at work in modern functional genomics
As the abundance and complexity of data in modern genomics is growing exponentially and
consistently exceeds the gains in computational power (as described by the Moore’s law),
there is a growing need for using smart computational strategies for data analysis. Our team
of interdisciplinary scientists is aiming to provide the scientists working on functional genomics
with computational methodology adequate for the challenges of the next-generation
sequencing.
In particular, we have worked on the problem of identification of functional regulatory
sequences in the genomes of different animals, from fruit flies, through laboratory mice to
humans. This problem requires effective methods for scanning through terabytes of
sequencing data to fish for regions enriched in specific sequence features, specific biochemical
contacts or both. It is usually done for the data, where the training data is limited and the
datasets are noisy, so any attempts of using out-of-box machine learning libraries is prone to
erroneous results. However, due to the fact that the data availability is growing very quickly,
there is certainly room for use of customised machine learning methods to arrive at reliable
conclusions based on cross-comparisons between datasets.
In my talk, I will describe some of our methodology, including our work on Dynamic Bayesian
Networks for the task of classification of regulatory regions. I will also discuss our experiences
in using other, more standard classification techniques like Random Forests or Support Vector
Machines and specific scenarios where it led to biologically relevant conclusions.
33
Notes
34
Notes