Research Overview

The primary research agenda of APL is to understand the organizing principles of adaptive biological systems, particularly those of the visual system. Our central hypothesis is that the visual system is organized to perform hierarchical Bayesian inference. We are pursuing this hypothesis through a set of computational and electrophysiological studies in three different key topical areas in vision: the inference of 3D surface structures (mid-level vision) in the early visual areas, the analysis of motion in the dorsal visual stream, and the contextual analysis of object forms in the ventral visual stream.

We believe that the visual hierarchy emerges from a variety of learning and adaptation processes that encode the statistical regularities in the visual patterns of natural scene in different behavioral contexts. To discover these statistical regularities, we have constructed databases of static 3D scenes and dynamic movies to study the statistical basis of perceptual inference and the nature of neural codes for higher order and abstract information. We are carrying out neurophysiological experiments using single-unit and array recording techniques to elucidate these higher order neural codes, to study the principles of cortical inference, and to understand the mechanics and dynamics of learning and adaptation.

We hope these research efforts will advance our fundamental understanding of the visual systems, contributing to the knowledge and technical foundation for the development of new therapeutic treatments and neural interface devices to restore vision or at least alleviate the suffering of the visually impaired. Understanding the computational principles and neural algorithms of the adaptive active visual systems is essential to building the next generation of intelligent machines and the path to knowing ourselves.