Computational Neuroscience Group

Please note: this group does not really exist anymore in this lab, and this page obsolete. See here for more recent information.

Our goal is to build mathematical models of brain function.

In particular we're interested in modelling the visual cortex using methods from statistical theory. The statistical theory is largely based on independent component analysis (ICA) that is a new version of factor analysis or principal component analysis. We use ICA to model the statistical structure of natural images, that is, the kind of images that we see in everyday life (or rather, we have seen during our evolutionary history and childhood). Then we investigate the links between this statistical structure and the human visual system. Applications of the results on image processing are investigated as well.

More background

Lisää tutkimuksen taustasta (in Finnish)

Members

Patrik Hoyer	Jarmo Hurri	Aapo Hyvärinen	Mika Inki
		(group leader)

Projects

Modelling complex cells and topography by extensions of ICA

Independent component features from colour and stereo image data

Statistical structure of complex cell outputs: emergence of contour coding

Statistics of natural video sequences

Image processing applications

Activities

Fourth International Workshop on Independent Component Analysis and Blind Source Separation (ICA2003).

NIPS2002 workshop on learning invariant representations.

First International Workshop on Generative-Model-Based Vision (GMBV 2002).

Special Issue on Independent Component Analysis and Blind Source Separation in Neurocomputing.

Special session on Models of natural image statistics at ICONIP-2000.

Representative publications

J. Hurri and A. Hyvärinen. Simple-Cell-Like Receptive Fields Maximize Temporal Coherence in Natural Video. Neural Computation, in press.
Postscript  /  gzipped PostScript  /  PDF

P.O. Hoyer and A. Hyvärinen. A multi-layer sparse coding network learns contour coding from natural images. Vision Research, 42(12):1593-1605, 2002.
Postscript  /  gzipped PostScript  /  PDF

A. Hyvärinen and P.O. Hoyer. A two-layer sparse coding model learns simple and complex cell receptive fields and topography from natural images. Vision Research, 41(18):2413-2423, 2001.
Postscript  /  gzipped PostScript  /  PDF

A. Hyvärinen and P.O. Hoyer. Emergence of phase and shift invariant features by decomposition of natural images into independent feature subspaces. Neural Computation, 12(7):1705-1720, 2000.
Postscript  /  gzipped PostScript  /  PDF

P.O. Hoyer and A. Hyvärinen. Independent Component Analysis Applied to Feature Extraction from Colour and Stereo Images. Network: Computation in Neural Systems, 11(3):191-210, 2000.
Postscript  /  gzipped PostScript  /  PDF

Related research

FastICA and other one-unit algorithms for ICA,
and the FastICA MATLAB Package.

Noisy ICA and applications to image denoising

This research group originated from ICA research.
See also the Bayes Group.

NeuroHUT