NIPS*98 Workshop

Statistical Theories of Cortical Function

Organizers: Rajesh P.N. Rao, Bruno A. Olshausen and Michael S. Lewicki

When: Friday, December 4, 1998

Where: Breckenridge, Colorado


Although a considerable amount of data now exists on cellular, physiological, and neuroanatomical properties of the cerebral cortex, it is only in recent years that attempts have been made to link these distinctive properties of the cortex to concrete computational principles. Especially promising results have been obtained using statistical principles such as maximum likelihood and Bayesian estimation to derive efficient algorithms for learning and perception. The application of these algorithms to natural signals has been shown to generate receptive field properties similar to those observed in the cortex. The instantiation of some of these algorithms in the form of "analysis-synthesis" loops has suggested possible functional roles for the reciprocal feedforward-feedback connections between cortical areas. The use of lateral connections in some of these models suggests analogous roles for the horizontal connections between neurons within the same cortical area.

The goal of this one-day workshop is to bring together researchers interested in exploring the use of well-defined statistical principles in understanding cortical structure and function. Topics that are expected to be addressed include: theories of perception and neural representations in the cortex, statistical interpretations of the function of lateral and cortico-cortical feedback connections, and development of cortical receptive field properties from natural signals. Speakers will be encouraged to suggest how their particular approaches may be implemented in the cortex and what predictions, if any, their model makes. After each brief session of talks, workshop participants will be given the opportunity to ask questions and to discuss the merits and weaknesses of each of the models. A final panel-style discussion will attempt to delineate the major issues of contention/controversy that arose during the talks/discussions, and will seek to enumerate the major challenges faced by future models of cortical function.


Schedule:

Morning Session: 7:30 - 10:30 AM

7:30 - 7:45 Christian Piepenbrock (TU Berlin): Lateral Cortical Competition and Orientation Selectivity

7:45 - 8:00 Richard Zemel (Arizona): A network model for the perception of multiple orientations

8:00 - 8:15 Alex Pouget (Georgetown): A Neural Theory of Ideal Observers

8:15 - 8:30 Eero Simoncelli (NYU): Are the Nonlinear Aspects of Striate Cortical Neurons Statistically Sensible?

8:30 - 8:45 Zhaoping Li (University College London): Intracortical lateral connections for pre-attentive segmentation

8:45 - 9:00 Break

9:00 - 9:15 Bill Freeman (MERL): Learning the joint statistics of visual images and their scene interpretations

9:15 - 9:30 Daniel Kersten (Minnesota): Perception as Bayesian Decision Theory: Discounting the Color of Mutual Illumination (joint work with Anya C. Hurlbert and Marina Bloj)

9:30 - 9:45 Yair Weiss (UC Berkeley): Bayesian combination of local motion signals in human vision (joint work with Ted Adelson)

9:45 - 10:00 Jong-Hoon Oh(Pohang University): Uppropagation Algorithm and Visual Data Processing

10:00 - 10:30 Open Discussion

------- Ski Break -------

Afternoon Session: 4:00 - 7:00 PM

4:00 - 4:15 Zoubin Ghahramani (University College London): Distributed Representations and Spike Timing (joint work with Geoffrey Hinton)

4:15 - 4:30 Michael Lewicki (Salk): Encoding time-varying signals using spike-like representations

4:30 - 4:45 Bruno Olshausen (UC Davis): Sparse codes and spikes

4:45 - 5:00 Rajesh Rao (Salk): Predictive coding and recurrent excitation in the neocortex

5:00 - 5:15 Sue Becker (McMaster): Modelling expectancy and translation-invariance in musical sequence processing (joint work with Kate Stevens and Laurel Trainor)

5:15 - 5:30 Break

5:30 - 5:45 Tony Bell (Interval): Information theory in the cortex?

5:45 - 6:00 Nathan Intrator (Tel Aviv): Neuronal goals and related coding

6:00 - 6:15 Alan Yuille (Smith-Kettlewell): From Generic to Specific: An Information Theoretic Perspective on Cortical Function

6:15 - 6:30 Jean-Pierre Nadal (Ecole Normale Supérieure): Information theory and cortex

6:30 - 7:00: Panel Discussion


For further information, contact:
Rajesh Rao
The Salk Institute, CNL & Sloan Ctr
10010 N. Torrey Pines Road
La Jolla, CA 92037

Phone: 858-453-4100 x1215
Fax: 858-587-0417
WWW: http://www.cnl.salk.edu/~rao/
e-mail: rao@salk.edu