Cosyne 2010 Workshops
March 1-2, 2010
Snow Bird, Utah
Linearity and its discontents -- Is there life in a post-STRF world?
Stephen David (Maryland)
Yale Cohen (Penn)
Jonathan Fritz (Maryland)
As a linear model, the spectro-temporal receptive field (STRF) has enjoyed some success in characterizing neural responses in early sensory cortical areas. At higher levels of the sensory processing hierarchy, the ability of the STRF to characterize neural response properties diminishes rapidly. However, despite its limitations and the numerous proposed extensions and alternatives to the STRF, no single approach has managed to replace it as the canonical model in the field of sensory neuroscience.
Attempts to improve the ability of the STRF to characterize the responses of neurons in both early and late sensory-processing areas fall into several different categories: (1) new basis sets to describe sensory inputs, (2) new nonlinearities to explicitly capture the transformations applied to these inputs, and (3) new cost functions to incorporate ideas from machine learning theory or experimentally-informed priors. This workshop will focus on three central questions. First, what is the enduring value of the STRF, and why is it still so prevalent? Second, how successful have second-generation models been at improving our understanding of sensory processing? Finally, and perhaps most importantly, are there fundamentally different alternatives that offer better ways to characterize sensory responses in the cortex?
In this workshop, we will bring together theoreticians and experimentalists from different sensory systems to discuss these central questions and explore how our collective experience can be usefully applied to future models of sensory processing.