Machine Learning Meetings and Events
Group Meetings: Group meetings are held Mondays from 11am to Noon (talk starts 11:10am) in D.L. Pratt 290C unless otherwise noted. Meetings are coordinated by Hugo Larochelle.
Tea Talks: Tea talks are held every Wednesday at 4:00pm in D.L. Pratt 290C. Talks should be simple, accessible, and not exceed 15 minutes. Speakers bring snacks, make tea, and provide a copy of the presented paper.
Group Meeting May 6, 2009: Separating sources and analyzing causality in brain waves.
- Speaker: Aapo Hyvärinen
- Abstract:
Blind source separation methods have been succesful in removing technical artifacts from EEG and MEG data ("brain waves"), but so far decomposition of on-going oscillatory activity in the brain has been less succesful. Here I present a source separation method, Fourier-ICA, which is based on making a short-time Fourier transform before application of ICA. Experiments show that it focuses more on the oscillatory brain activity and ignores artifacts, where as basic ICA focuses on the artifacts and almost ignores brain activity. The method can also be analyzed in the case of kurtosis-based ICA; the analysis shows that the method can even separate gaussian time-dependent sources like "second-order" source separation methods. Next, I consider the question of causal discovery in this kind of data. Causal analysis can be performed by structural equation models, and we have shown that serious identifiability problem is such models can be solved by using the assumption of non-Gaussianity, like in ICA. I will show some very preliminary results on such analysis of MEG data.
- Notes: Wed 4pm in PT266