Presentation on theme: "MMACADAMIAMMa Macadamia Programme : Introduction to Research at the Department of Information and Computer Science (ICS) Erkki Oja Distinguished Aalto."— Presentation transcript:
MMACADAMIAMMa Macadamia Programme : Introduction to Research at the Department of Information and Computer Science (ICS) Erkki Oja Distinguished Aalto Professor, Director of Macadamia
Research at ICS Department: Macadamia research is coordinated by two national Centers of Excellence: Computational Inference Research Centre (COIN) Algorithmic Data Analysis Research Centre (Algodan); Pattern Discovery subgroup
Computational Inference Research Centre (COIN): : Finnish Centre of Excellence in Research, (also , , ) Highly competitive position (Aalto has 7 CoE’s in total in all research fields) Financed by Aalto, Academy of Finland, Orion Co. and Nokia Co.
COIN Centre of Excellence in Computational Inference: Introduction Erkki Oja, Director of COIN
Core Methodological Challenges C1: Learning models from massive data C3: Statistical inference in structured stochastic models C4: Extreme Inference engine C2: Learning from multiple data sources Future Computational Inference F1: Intelligent information access F2: Computational molecular biology and medicine Kaski Oja Myllymäki Niemelä Corander Aurell Laaksonen
F1: Intelligent Information Access
F2: Computational Molecular Biology and Medicine Fig.. Eye diagram showing the relationships between top 10 CCA coponents (middle), all Volsurf descriptors (left), and top genesets (right). CCA components are shown as circles, with numbers indicating the decreasing order of canonical correlation and letters indicating the sub-components. The widths of the curves from the components to Volsurf descriptors and genesets indicate the strength of the corresponding associations. For Volsurf descriptors the sub-component specific activity is shown, whereas for the genesets we show the overall activity of the component.
Computational X Climate Neuroscience History
SOM ICA Probabilistic Latent Variable Models, Bayes Reliability Linear mixtures FastICA Visualization, nonlin. dim. reduction Bayes blocks Nonl.dynamics, subspaces Nonlinear, non-neg. BSS Sparse PCA, DSS, nonneg. projections Relational models Relevance by data fusion Algorithmic research in Macadamia
Algorithmic work example: ICA ISI citations: 2230
Raha-automaattiyhdistys, RAY Fujitsu Services Fazer i-Sieve (Greece) Lingsoft Sanako Xerox Inria
ABB Teollisuuden Voima (TVO) Neste Oil Oyj UPM Kymmene Danisco Sweeteners Oy Keskuslaboratorio (KCL) Rautaruukki Raahe Steel
Important on-going and recent projects Emime, EU, personalized speech-to-speech translation PinView, EU, personal information navigator UI-ART, Aalto, multimodal augmented reality T4ME, EU NoE, technologies for multilingual Europe Pascal2, EU NoE, pattern analysis, stochastic modelling NOVAC, Academy, computational climatology MultiBio, Tekes, fusion of biological information sources Kulta, Tekes, modelling consumer behaviour
M.Sc. Theses Some Master of Science Thesis topics of previous Macadamia students -Modeling and profiling people’s way of living: a data mining approach to a health survey (Luis Gabriel de Alba Rivera) -Histogram equalization for noise robust speech recognition (Stevan Keraudy) -A proactive interface for image retrieval (Laszlo Kozma) -Fast variable selection using delta test (Dusan Sovilj)
M.Sc. Theses -Reinforcement learning in real time strategy games (Antonio Cocco Martins Gusmao) -Traffic flow simulation and optimization using evolutionary strategies (Alejandro Lopez Vidal) -Applying data mining for fault detection in energy data management (Yao Lin) -Mixture modelling of multiresolution 0-1 data (Prem Raj Adhikari) -…. and many more !!