Intelligent Signal Processing Jan Larsen Informatics and - PowerPoint PPT Presentation

Informatics and Mathematical Modelling / Intelligent Signal Processing Intelligent Signal Processing Jan Larsen Informatics and Mathematical Modelling Technical University of Denmark Web: isp.imm.dtu.dk Jan Larsen 1

Informatics and Mathematical Modelling / Intelligent Signal Processing DTU Facts � Independent state university since January 1st, 2001 � Annual budget 1382 million DKK or 186 million Euro � External research finances account for 35% � 1027 faculty/academic staff members, 948 technical and administrative staff � Organized in 15 departments � 1500 B.Sc., 4300 M.Sc., 500 Ph.D., and 500 international students � 380 courses in English, 10 international M.Sc. programs Jan Larsen 2

Informatics and Mathematical Modelling / Intelligent Signal Processing DTU Facts � Located in Lyngby 10km (6.3miles) north of Copenhagen � More than 80 buildings � Total floorage of about 375,000 m 2 (3.3million ft 2 ) � Campus size 1500m from North to South, 850m from East to West (0.9 by 0.5 miles) Jan Larsen 3

Informatics and Mathematical Modelling / Intelligent Signal Processing IMM � 86 faculty members � 32 administrative staff members � 60 Ph.D. students � 70 M.Sc. students annually � 6000 students follow an IMM course annually Jan Larsen 4

Informatics and Mathematical Modelling / Intelligent Signal Processing Competences and Research at IMM image processing and graphical communication intelligent signal processing integrated circuits and digital system operations research data safety numerical analysis knowledge-based systems geoinformatics software engineering mathematical statistics networks and distributed systems mathematical physics embedded systems programming languages Jan Larsen 5

Informatics and Mathematical Modelling / Intelligent Signal Processing ISP Group Activities Multimedia Humanitarian Neuroscience Demining Methods Algorithms •3 faculty •3 faculty •3 post docs •3 post docs •11 Ph.D. •11 Ph.D. students students Monitor •10 M.Sc. •10 M.Sc. Biomedical Systems students students Jan Larsen 6

Informatics and Mathematical Modelling / Intelligent Signal Processing Research Projects 1991-2001 CONNECT: fundamental research in neural networks and adaptive � systems 1997-2001 DMM: Distributed Multi Media research program � 1998-2001 THOR Center for Neuroinformatics � 1996-1999 Interdisciplinary Neuroscience Initiative - neuroimaging � collaboration in Copenhagen 1999-2002 SITE: signal and image processing methods for telemedicine � 2002-2006 International Center for Biomedical Research, minor contribution to � EEG/pain studies. 1997-2004 HBP: Human Brian Project - tools for analysis of brain scans � 2000-2003 MAPAWARMO - EU5FP project on mapping visual cortical regions in � awake, behaving monkey using functional MRI 2002-2005 AE-WATT - EU5FP project on neural network and adaptive � processing for monitoring and control of large diesel engines 1997- Humanitarian Demining Laboratory (HDL) � Jan Larsen 7

Informatics and Mathematical Modelling / Intelligent Signal Processing Network � IEEE Neural Networks for Signal Processing technical committee � Copenhagen Signal and Image Processing Graduate school (CISP) � Copenhagen Brain Research Center � Nordic Demining Research Forum � European partners through more 6th EU FP LoI’s Jan Larsen 8

Informatics and Mathematical Modelling / Intelligent Signal Processing How do we colaborate � Master thesis projects � Ph.D. projects funded by DTU � Industrial Ph.D. projects � Research projects funded by Danish research councils / EU � Commissioned research and development � Special designed competence development courses Jan Larsen 9

Informatics and Mathematical Modelling / Intelligent Signal Processing Competences in signal processing pattern recognition data and webmining statistical Learning nonlinear adaptive machine learning bayesian Jan Larsen 10

Informatics and Mathematical Modelling / Intelligent Signal Processing Methods and Algorithms Theory Models � Bayesian learning � Neural networks � Mean field methods � Relevance vector machines � Unsupervised/supervised and � Independent Component combined learning Analysis � Classification/regression � Time-series models � Model selection � Graphical Models � Nonstationary modeling � Gaussian Processes � Datafusion � Mixture models � Generalization � Outlier detection � Robust modeling � Active learning Jan Larsen 11

Informatics and Mathematical Modelling / Intelligent Signal Processing Our modeling approach Prior knowledge Model Selection and data Validation Learning Jan Larsen 12

Informatics and Mathematical Modelling / Intelligent Signal Processing Learning Based Modeling p( y | x w , ) ⋅ p x ( | w ) u u Labels Measurements Parameters Unsupervised Supervised Learning joint probability Learning Learning Jan Larsen 13

Informatics and Mathematical Modelling / Intelligent Signal Processing Supervised Learning � Selection of model family � Model structure optimization � Preprocessing � Performance issues – Generalization � A priori knowledge constraints/probabilistic – Confidence – ROC curves � Learning (ML,MAP,Bayes) – Confusion matrices � Active learning – Reproducibility � Outlier detection � Visualization/interpretation – Erroneous label – Importance analysis – Unknown label p y ( | x w , ) s Jan Larsen 14

Informatics and Mathematical Modelling / Intelligent Signal Processing Unsupervised Learning � Identification of structure, � Novelty detection – e.g. clusters identification of new clusters � Can be used as � Hierarchical clustering preprocessing in combination with supervised learning � Methods – ICA – Probabilistic clustering using RBF neural networks p x ( | w ) u Jan Larsen 15

Informatics and Mathematical Modelling / Intelligent Signal Processing Neuroscience � Analysis of EEG, PET, MRI, fMRI images � Lyngby Toolbox � Web repository hendrix.imm.dtu.dk Programs: CBRC, HBP, EU-BIOMED II, EU-TMR, EU, Res.Counc., INC, NIH Jan Larsen 16

Informatics and Mathematical Modelling / Intelligent Signal Processing Multimedia � Adaptive tools for shared virtual environments � Real time multi-modal communication � Tools for intelligent mobile phones (CDMA) � Sound processing � Webmining (content,structure,usage) Programs: Center for Multimedia, DMM Jan Larsen 17

Informatics and Mathematical Modelling / Intelligent Signal Processing Humanitarian Demining � Humanitarian Demining Lab (Ørsted•DTU,IMM,FOFT) – Detection and clearance of mines – Quality control and clearance validation – Socio-economical decision support systems – Resource management systems � Nordic Demining Research Forum � New EU 6FP EoIs Jan Larsen 18

Informatics and Mathematical Modelling / Intelligent Signal Processing Monitor Systems � Supervision and fault- diagnosis of marine engines � Neural optimization of pulse plating in print circuit boards � Monitoring of power distribution networks � Multi-channel adaptive analysis of revolving machines � Adaptation of mobile base station filters Program: AE-WATT Jan Larsen 19

Informatics and Mathematical Modelling / Intelligent Signal Processing Biomedical � Sun exposure modeling 0.045 F � Skin cancer diagnosis 0.04 � Bioinformatics 0.035 B A G Sensitivity E 0.03 D B C 0.025 0.02 Program: Signal and 0.015 Image Processing for 0.01 3500 3000 2500 2000 1500 1000 500 wavenumber/cm -1 Telemedicine Jan Larsen 20