SNN / Technologiestichting STW
| Programma- en abstract boek
Lerende Oplossingen
22 oktober, 2003 Auditorium Radboud Universiteit Nijmegen
2003
symposium Lerende Oplossingen op 22 oktober 2003 in Nijmegen zijn gepresenteerd.
Organisatoren
+31 (0)24 354 14 35
Dr. H.J. Kappen | Radboud Universiteit Nijmegen/SNN Dr.ir. R.P.W. Duin | Technische Universiteit Delft Dr.ir. B.J.A. Kröse | Universiteit van Amsterdam Dr. W. Segeth | Technologiestichting STW A. Wanders | SNN
Plaats Congres Auditorium en Kasteel Heyendael | Radboud Universiteit Nijmegen | Geert Grooteplein 15 | Postbus 364 | 6500 AJ Nijmegen | tel portier +31 24 3610 112 | mobiel SNN 06 2052 6691
| fax
52 6691
Dit boek is een bundeling van abstracts van de posters welke tijdens het ééndags
Contactadres SNN | t.a.v. A. Wanders | Postbus 9101 | 6500 HB Nijmegen | | tel +31 24 361 42 45 | fax +31 24 354 14 35 | e-mail
[email protected] | website www.snn.kun.nl/nederland/learning
Dit symposium wordt gesteund door Technologiestichting STW | Dr. W. Segeth |
[email protected] SNN | Dr. H.J. Kappen |
[email protected]
Lerende Oplossingen 2003 Bayesiaanse methodes in theorie en praktijk
Welkom op het symposium Lerende Oplossingen. Dit is de derde maal dat wij dit symposium organiseren, de twee voorafgaande keren waren in 2000 en in 2002. Het doel van het symposium is om op het onderzoeksgebied dat we aanduiden met Adaptieve Intelligentie een actueel overzicht te geven van de stand van de techniek met een speciale aandacht voor de toepasbaarheid in de praktijk. Het symposium beoogt een ontmoetingsplaats te zijn tussen onderzoekers en toepassers. Wat is Adaptieve Intelligentie en hoe verhoudt het zich tot neurale netwerken? De SNN is oorspronkelijk opgezet met als doel het stimuleren van onderzoek op het gebied van neurale netwerken. Echter, in de laatste 10 jaar is het vakgebied sterk verbreed. Zo is het nu mogelijk om lerende methodes te combineren met regelgebaseerde methodes en is de robuustheid van de oplossingen sterk verbeterd door gebruik te maken van statistische technieken. Het formalisme van de Bayesiaanse statistiek heeft een belangrijke bijdrage geleverd aan de integratie van deze drie traditioneel sterk verschillende vakgebieden. Lerende methodes zoals neurale netwerken zijn in het algemeen dus slechts een deel van een totaal oplossing en met Adaptieve Intelligentie bedoelen we dit bredere gebied. In 1995 is de Technologiestichting STW - in nauwe samenwerking met de Stichting Neurale Netwerken, SNN begonnen met de uitvoering van het onderzoeksprogramma Neurale Netwerken/Adaptieve Intelligentie. De eerste drie oproepen voor onderzoeksvoorstellen (in 1995, 1997 en 1999) resulteerden elk in een vijftal gehonoreerde projecten. De vierde oproep is in 2002 uitgegaan en tegen het eind van 2003 zal bekend worden welke projecten gehonoreerd worden. Tijdens de middag zal een overzicht gegeven worden van lopende en recent afgeronde STW-SNN projecten. Daarnaast zal ook overig onderzoek worden gepresenteerd zodat een vrijwel volledig overzicht ontstaat van het Nederlandse onderzoek op dit gebied. Tevens zijn we erg verheugd, dat we Philips Research en Microsoft Research bereid hebben gevonden om een overzicht geven van hun activiteiten op dit gebied. In de ochtend zijn er twee parallelle tracks. Er is een industriële track waar drie bedrijven hun lerende oplossingen zullen presenteren. Voor de meer theoretisch geïnteresseerden is er een tutorial van Micheal Kearns over speltheorie. Deze tutorial is tevens onderdeel van de Belgisch Nederlandse AI Conferentie (BNAIC 2003) die op 23 en 24 oktober in Nijmegen wordt gehouden. Wij wensen u een leerzame dag toe. Bert Kappen Wouter Segeth
Programma Lerende Oplossingen 2003 09:30 Registratie en koffie in AUDITORIUM OCHTEND A 10:00 Opening, Dr. H.J. Kappen, KU Nijmegen 10.15-12.15 Adaptive Intelligence in practice 10.15 New Concepts in Knowledge Management Ivo de Blinde en Edgar van Oostrum, Knowlutions, SPSS Are you ready to know More! In these challenging times the traditional indicators time, quality and cost are no longer sufficient to assess the performance of companies. Flexibility, adaptation and learning capabilities play a vital role in business processes. This is why successful companies regard knowledge management as one of the key success factors in the implementation of new business strategies. Proper management of the knowledge base for value creation can provide a window of opportunity for the improvement of your companies R&D and problem solving capabilities. Knowlution BV and SPSS Inc. have combined unmatched technologies in a new approach to capture, manage and support knowledge distribution, application and creation.
10.50 Predicting Consumer behavior Tom Heskes, SMART Research B.V. In this presentation, I will focus on a specific forecasting problem: predicting the sales of newspapers. The most intriguing aspect is that one has to generate predictions for a huge number of similar time series on a daily basis. I will illustrate how we exploit these similarities to have the different outlets "learn from each other". Studies for several European publishers reveal consistent and very significant performance improvements.
11.25-11.40 Koffie 11.40 Predicting tourist travel behavior Christoph Engels, Thinking Networks Germany Planning and prediction of tourist travel behavior The tour operator business has changed dramatically over the last dekade. While in the early ninties it was characterized as a seller market it had changed to a buyer market. Tour operators have tried to face this development by introducing sophisticated planning and predition systems to support the concept of revenue management. During the last six years Thinking Networks has implemented the first revenue management system for tour operators worldwide at world largest tour operator TUI. The talk will describe the different prediction tasks and our approaches for strategic marketing planning, demand prediction for planning and short-term prediction for last minute and discount offers. The used methods include neural networks and hard modelling for prediction and text mining tasks.
OCHTEND B 10.15-12.15 Tutorial Computational Game Theory 11.15 tot 11.30 Koffiepauze Prof. M. Kearns, University of Pennsylvania Recently there has been renewed interest in game theory in several research disciplines, with its uses ranging from the modeling of evolution to the design of distributed protocols. In the AI community, game theory is emerging as the dominant formalism for studying strategic and cooperative interaction in multi-agent systems. Classical work provides rich mathematical foundations and equilibrium concepts, but relatively little in the way of computational and representational insights that would allow game theory to scale up to large, complex systems. The rapidly emerging field of computational game theory is addressing such algorithmic issues, and this tutorial will provide a survey of developments so far. The tutorial will be self-contained, assuming no prior knowledge of game theory.
MIDDAGPROGRAMMA in AUDITORIUM 12:15 Lunch 13:30 Statistical media processing at microsoft research Prof. J. Platt, Microsoft Research in Redmond Statistical Media Processing is the intersection between media (such as image, video, music, speech) and machine learning (statistical algorithms). Our research in this area has produced practical technology that can improve products. For example, we have created a clustering algorithm for automatically organizing digital photos; an algorithm to learn similarities between songs, to automatically generate music playlists; and an algorithm to automatically extract noise-robust features from audio, to identify music files or streams.
14:15 Overview Adaptive Intelligence in The Netherlands Dr. H.J. Kappen, KU Nijmegen 14:45 Postersessie 16:00 Algorithms in Ambient Intelligence Emile Aarts,Philips Research and Eindhoven University of Technology In the near future our homes will have a distributed network of intelligent devices that provides us with information, communication, and entertainment. Furthermore, these systems will adapt themselves to the user and even anticipate on user needs. These consumer systems will differ substantially from contemporary equipment through their appearance in peoples’ environments, and through the way users interact with them. Ambient Intelligence is the term that is used to denote this new paradigm for in-home computing and entertainment. Salient features of this new concept are ubiquitous computing, and natural interaction. Recent developments in technology, the Internet, the consumer electronics market, and social developments indicate that this dream might become reality soon. First prototypes of ambient intelligent home systems have been developed, but the realization of true ambient intelligence calls for much additional research of multidisciplinary teams consisting of technologists, designers, and human behavior scientists. Algorithms play a central role in the development of ambient intelligence. Key features such as quality of service, load balancing, context awareness, personalization, adaptation, and anticipatory behavior can be realized through sophisticated on-line algorithms that run in real-time. The presentation will outline the current status of the development of algorithms for ambient intelligence.
16:40 Postersessie en receptie 18.00 Sluiting
Index STW Projecten
Page
Poster/ Demo
I
Posterpresentaties |
Application of Cluster Variation method to Genetic Linkage Analysis Kees Albers, Bert Kappen
|
A Virtual Mobile Robot Left Ventricle Wall Delineation in Cardiac MR Images F. Behloul, B.P.F. Lelieveldt, R.J. van der Geest, J.H.C. Reiber
1
P15
2
P9
|
A Dynamic Bayesian Network for Polyphonic Music Transcription Ali Taylan Cemgil, Bert Kappen
3
P8
|
Perceptual Grouping Remco Duits
5
P21
|
Computer-Aided Detection of Breast Cancer in Mammograms S. van Engeland, S.Timp, C. Varela, N. Karssemeijer, C.C.A.M. Gielen
7
P17
|
Learning Concepts in Real World Embodied Agents Stephan ten Hagen, Bram Bakker, Ben Kröse
8
P3
|
Online Auction Bidding Strategies for Logistics P.J. 't Hoen, J.A. La Poutré A Simple Multi-Agent System for Evaluating Strategies Th. M. Hupkens
10
P7
11
P13
| |
A Symbolic Approach to Music Recognition Nico Jacobs, Filip Van den Borre, Lennert Smeets, Evarest Schoofs, Hendrik Blockeel
12
D5
|
Preprocessing Documents to Answer Dutch Questions Valentin Jijkoun, Gilad Mishne, Maarten de Rijke
13
P22
|
ProAnita: A Multi-Agent Solution for Legitimate Information Retrieval Femke de Jonge, Nico Roos, Pieter Spronck, Steven de Jong
14
D6
|
Sample Complexity for One-Class Classifiers Piotr Juszczak, Robert P.R.Duin
15
P25
|
Robust Active Fault-Tolerant Control Stoyan Kanev, Michel Verhaegen, Alexander Efremov
16
P11
|
Promedas: A Diagnostic Decision Support System Bert Kappen, Wim Wiegerinck, Ender Akay, Marcel Nijman, Jan Neijt, André van Beek
17
D1
|
Icon Based Communication in an International Environment P.A. Kersseboom, Ir. B.Bruggeman, Dr. L. Rothkrantz
19
P12
|
Muli-Robot Decision Making Using Coordination Graphs J.R. Kok, M.T.J Spaan, N. Vlassis
21
P2
|
Probabilistic Methods for Localizing Lino, the User-Interface Robot B.J.A. Kröse, J.M. Porta
23
P6
|
Searching Maps by CNNs on an FPGA Suleyman Malki, Lambert Spaanenburg
25
P19
|
Multi-View Active Appearance Models: Application to X-ray LV Angiography and Cardiac MRI C.R Oost, M. Üzümcü, D. Kaandorp, J.H.C. Reiber, M. Sonka, B.P.F. Lelieveldt
26
P10
|
Sales Forecasting Through Aggregation: Integrating Neural-Bayesian and Knowledge Based Methods Pim Ouwehand
28
P16
|
Speeding up the Classification of Hyper-Spectral Data Using the Dissimilarity Representation Pavel Paclik, Robert P.W. Duin, Serguei Verzakov
29
P26
|
On (not) Making Dissimilarities Euclidean Elzbieta Pekalska, Robert P.W. Duin
31
P22
|
Robust Manifold Learning Dick de Ridder, Robert P.W. Duin, Vojtech Franc
32
P23
|
Combining Autofluorescence and Diffuse Reflectance Spectroscopy for Improving Lesion Diagnostics in Lungs M. Skurichina, A. Amelink, M. Bard, H. Sterenborg, R. Duin
33
P24
|
Optimizing Single-Copy Newspaper Sales with JED Jan-Joost Spanjers, Marco Bloemendaal, Tom Heskes
34
D2
|
Online Adaptation of Computer Game Opponent AI Pieter Spronck, Ida Sprinkhuizen-Kuyper, Eric Postma
36
P1
|
The Performance of (Least Squares) Support Vector Machines for Multivariate Spectral Calibration U. Thissen, B. Üstün, W.J. Melssen, L.M.C. Buydens
37
P18
|
A Constrained EM Algorithm for Large-Scale Mixture Modeling J.J. Verbeek, N. Vlassis, J.R.T.J. Nunnink Intelligent Traffic Light Control Marco Wiering
38
P4
40
D3
| |
The Xetal INCA: an Embedded Platform for Combining Classification Algorithms and Conventional Image Processing Rik Wetzels, Philips CFT
D4
|
Iterated Extended Kalman Smoothing with Expectation-Propagation Alexander Ypma, Tom Heskes
42
P14
|
Probabilistic Models for Distributed Surveillance Wojciech Zajdel, Ben Kröse
44
P5
|
Maturing a Network Structure for Rule Extraction Berend Jan van der Zwaag, Suleyman Malki and Lambert Spaaneneburg
46
P20
STW Projecten gehonoreerd in 1997 NGN 4480: Knowledge Representation with Neural Networks Postertitel: Application of Cluster Variation method to Genetic Linkage Analysis Kees Albers, Bert Kappen
1
UGN 4496: Neural Network Approach to Scale Space Grouping in Image Analysis Postertitel: Perceptual Grouping 5 Remco Duits NIF4494: Quantization of Temporal Patterns by Neural Network Postertitel: A Dynamic Bayesian Network for Polyphonic Music Transcription 3 Ali Taylan Cemgil, Bert Kappen NCH 4501: Monitoring, Identification and Control of Chemical Systems and Processes using Neural Networks Postertitel: The performance of (Least Squares) Support Vector Machines for multivariate spectral calibration U. Thissen, B. Üstün, W.J. Melssen, L.M.C. Buydens
37
LGN 4503: Application of a fuzzy neural network to medical imaging segmentation Postertitel: A Virtual Mobile Robot Left Ventricle Wall Delineation in cardiac MR Images 2 F. Behloul, B.P.F. Lelieveldt, R.J. van der Geest, J.H.C. Reiber DEL 4506: Neuro-Fuzzy Modeling in Model Based Fault Detection, Fault Isolation and Controller Postertitel: Robust Active Fault-Tolerant Control 16 Stoyan Kanev, Michel Verhaegen and Alexander Efremov
P15
P21
P8
P18
P9
P11
STW Projecten gehonoreerd in 1999 NNN 5322: A decision support system for medical diagnosis using a large probalistic network Postertitel: Promedas: A diagnostic decision support system Bert Kappen, Wim Wiegerinck, Ender Akay, Marcel Nijman, Jan Neijt, André van Beek ENN 5323: Sales forecasting through aggregation Postertitel: Neural-Bayesian and knowledge based methods Pim Ouwehand ANN 5312: Probabilistische modellen voor het volgen van meerdere objecten met meerdere sensoren Postertitel: Probabilistic Models for Distributed Surveillance Wojciech Zajdel, Ben Kröse RNN 5316: Improving endoscopic detection of lungcancer using autofluorescence spectroscopy Postertitel: Combining autofluorescence and diffuse reflectance spectroscopy for improving lesion diagnostics in lungs M.Skurichina, A. Amelink, M. Bard, H. Sterenborg, R. Duin NNN 5321: Graphical models for data mining Postertitel: Iterated extended Kalman smoothing with expectation-propagation Alexander Ypma, Tom Heskes
17
D1
28
P16
44
P5
33
P24
42
P14
Posterpresentaties
Plattegrond Posters en Demonstraties
SPSS
Elsevier
Registration
D1 Kappen
D2 Spanjers
D3 Wiering
D4 Wetzels
D5 Jacobs
D6 De Jonge
P9
P14
P19
P20
P21
Behloul
Ypma
Malki
Zwaag
Duits
P10
P15
P22
P23
Oost
Albers
Jijkoun
Ridder
P11
P16
Kanev
Ouwehand
P24 Skurichina
P12
P17
P25
P26
P26
Kersseboom
Engeland
Juszczak
Pavlik
Pekalska
P13
P18
Hupkens
Thissen
P1 Spronckl
P2
P3
P4
P5
P6
P7
P8
Kok
Hagen
Verbeek
Zajdel
Kröse
Hoen
Cemgil
