Introduction

This is the first 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 Chair (Deniz Erdogmus, derdogmus@ieee.org).

Benchmark Problems

This year’s 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.
  1. Contribution by Andrzej Cichocki: Evaluation and development of blind source separation algorithms in linear instantaneous mixtures for large-scale (high-dimensional), ill-conditioned, and noisy measurements.

  2. Contribution by Vince Calhoun: Evaluation and development of intelligent artifact rejection in multidimensional signal processing, posed as a problem of eye-artifact removal in the context of event-related potential detection from EEG measurements.

  3. Contribution by Kenneth Hild: Evaluation and development of noise reduction algorithms and methodologies, posed as a problem of signal denoising to reduce number of required trials for MEG measurements.
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 August 1st, 2005 in the form of data files, figures/text, and Matlab code (as required).

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

Awards will also be presented to the winners at the meeting. The winners in each category will be allocated time at the MLSP 2005 meeting to present their methodology to the attendees and other competition participants. 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 short or long papers to a journal special issue that will be organized following the conference.

Good Luck, MLSP Organization Committee