Participants included: M. Anderson , D. Ascher, Amir Assadi , Hagai Attiar, Carlos Brody , Peter Dayan, Adam Elga , Winfried Fellenz , Jean-Marc Fellous , David Field, Jozsef Fiser , F. Frasso , B. Girish , J. Kralik, Chris Lee , Yephi Lin, B. Mensh , Eli Nelken, Bruno Olshausen , Nestor Parga, Bijan Pesaran , Bill Press , Avi Rabinavitch , Pam Reinagel , Cooper Roddey , Dan Ruderman, Emilio Salinas, Yair Weiss.
This CNS96 workshop met on the afternoon of Tuesday, July 16, to discuss the statistics of images/sound and their relation to sensory systems. We started with David Field and Dan Ruderman giving a background of work that has been done in the field over the past 40 years. A major result in this period was the discovery of a 1/f amplitude spectrum in natural images. Dan Ruderman then described some recent work he has done looking at the principal components of the L,M, and S cone activities in response to images, showing that these result in separate luminance and color opponency channels. David Field then discussed some of the properties that one might expect to be common across sensory modalities, for example correlated inputs, multi-scale structure (e.g., in audition, large objects resonating vs. small objects), intensity histograms that are log-normal, and localized events. There was some discussion of the difficulties posed in considering aspects of the stimulus that are important for behavior, or survival, as opposed to those statistics that are simply "there" in the environment. For example, in the auditory domain, a collection of natural stimuli may typically be overwhelmed with animal calls, which are behaviorally relevant. Does this pose a problem for the statistical approach? (not resolved) We ended by offering a wish list of problems we hope to see progress on over the coming years:
Back to Workshop Summaries page
Back to CNS*96 Home Page.