MLSP 2006


The extended submission deadline for the competition is 14 April 2006.

This is the second data analysis and signal processing competition that is being organized in conjunction with the IEEE Machine Learning for Signal Processing Workshops. This competition is envisioned to become an annual event where problems relevant to the mission and interests of the MLSP community will be presented with the goal of advancing the current state-of-the-art in both theoretical and practical aspects. The problems are selected to reflect the current trends to evaluate existing approaches on common benchmarks as well as areas where crucial developments are thought to be necessary. If you would like to contribute a problem for next year’s MLSP competition please contact the Competition Chairs Deniz Erdogmus and Vince Calhoun.

Benchmark Problems

The competition consists of three problems. All three problems are of general interest to the signal processing community and bear strong importance and relevance to the advancement of the current state-of-the-art in the application of advanced adaptation and machine learning concepts to signal processing problems.

Problem 1
Evaluation of ICA algorithms for large-scale, ill-conditioned, and noisy mixtures. Contribution by Andrzej Cichocki and Deniz Erdogmus.
Problem 2
Processing of fMRI data for an auditory oddball task. Contribution by Vince Calhoun.
Problem 3
Denoising of MEG measurements. Contribution by Kenneth Hild.

Participation Procedures

Please download the pdf files linked through the contributor names that describe the benchmark problems in detail and point to the websites where you can download the competition datasets. The participants should report their findings via email to DERDOGMUS@IEEE.ORG by March 31st, 2006 in the form of documents and Matlab codes.


The contributors of the winning entries for each problem will be invited to submit papers to the conference describing their approach and to present their results at the competition special session. These lectures will allow participants to discuss their experiences, share ideas, and shape future directions of research to pursue. There will also be an opportunity to contribute papers to a journal special issue that will be organized following the conference.

The winners of the 2005 Data Analysis of Competition:

Problem 1
ICA Benchmark: S.C. Douglas, J. Chao
Problem 2
Artifact Reduction in Multichannel EEG: N. Nicolaou, S. Nasuto
Problem 3
Single-trial MEG Denoising: M. Grosse-Wentrup and M. Buss

Good Luck,
MLSP Organization Committee