G80.3042/004, Spring Semester, 2002

Statistical Analysis and Modeling of Neural Systems



Instructor: Eero Simoncelli
Co-organizer: Liam Paninski
Lectures: Tuesdays 11-1, Meyer 813
Pre-requisites: Mathematical Tools for Neural Science, or equivalent.

Brief Description: A seminar-style course, focusing on use of statistical techniques for modeling and analyzing neural data. The course will be divided into two or three segments, each focusing on a different subtopic. Each segment will begin with a lecture providing history, background material, and some mathematical framework for the associated topic. Subsequent classes will consist of presentation and discussion of journal papers on the associated topic.

Ongoing seminar schedule

First topic: The efficient coding hypothesis

Connections to information theory. Extensions to adaptation.


Date Presenter Title
Jan 29 Eero Simoncelli Introduction. See Barlow, `01, Simoncelli and Olshausen, `01, and Tishby, Pereira, and Bialek `00.
Feb 5 Liam Paninski and Odelia Schwartz Linear models / ICA. See Hyvarinen and Oja `99 (tutorial) and Bell and Sejnowski `97.
Feb 12 Jonathan Pillow and Dajun Xing Some generalizations. See Hyvarinen and Hoyer `00, and Baddeley, `96, van Hateren and van der Schaaf `98.
Feb 19 Jenny Li and Jonathan Pillow Experimental comparisons. See Baddeley et al. `97, Treves et al,`99, Nirenberg et al `01, reply by Meister `01, reply to the reply, etc., Rieke et al `95, and Attias and Schreiner, `97


Second topic: Static estimation and decisions with neural populations

Standard decision-theoretic setup with (dependent) neural activity as ``data''. List of references here.

Date Presenter Title
Feb 26 Eero Simoncelli and Liam Paninski Introduction. See also, e.g., Seung and Sompolinsky, `93, Britten et al. `92.
Mar 5 Nicole Rust and Mehrdad Jazayeri Seung+Sompo93, pouget_etal99
Mar 12 no class spring break
Mar 19 David Hammond and Jenny Li zhang/sejn99, Eurich et al.
Mar 26 Dajun Xing and Odelia Schwartz Brunel and Nadal, `98, Shadlen et al, `96 (a scanned copy of the Shadlen paper is available here.)


Third topic: Reading dynamic information in spike trains

Point processes. Stimulus reconstruction techniques, both linear (Wiener, Kalman filters) and nonlinear (Bayesian, ``particle'' filters). Bremaud, Bialek, Peskin, Brown, Gao, Zhang.

Date Presenter Title
April 2 Liam Paninski and Jenny Li Introduction via linear filtering. Warland et al, `97
April 9 Dajun Xing and Jonathan Pillow Bayesian approaches. Zhang et al, `98, Brown et al, `98, and Brown et al, `01
April 16 David Hammond and Jenny Li More interesting spiking models. Kass and Ventura, `01, Brown et al, `02
April 23 Odelia Schwartz and Eero Simoncelli Simoncelli et al. `02
May 1 Simon Schultz Studying the information content of spike trains; note special day (wed instead of tues)



Site maintained by liam@cns.nyu.edu