Our laboratory has invented experimental and analytical tools to address these obstacles and decipher ventral pathway information processing at a quantitative, neural coding level. We use shape morphing techniques and genetic algorithms to guide stimulus sampling in the virtual infinity of shape space based on neural response levels. We are currently experimenting with genetic algorithms guided by large populations of neurons sampled over long time periods with chronic microwire array recording.
To analyze ventral pathway responses, we parameterize object shapes as clouds of dense samples in multi-dimensional contour, surface, and/or medial axis geometric space. These parameterizations are used to model neural responses as linear/nonlinear functions based on multi-dimensional Gaussian tuning regions. We are beginning to study how the ventral pathway represents physical properties (e.g. distribution of mass, flexibility, articulation) as well as shape. In collaboration with Dr. Kristina Nielsen, we are beginning to use 2-photon imaging and array recording to study how local circuits transform local shape information into larger constructs. Our laboratory is dedicated to developing new experimental/computational methods for collecting and decoding large-scale ventral pathway data and its meaning in relationship to the real world.