Mechanisms of Perceptual Learning: Data and Model

 

Mechanisms of Perceptual Learning: Data and Model

Ricketts 211

* How does the brain adapt to an ever-changing environment? How does it strike a balance between continuity and change? What neural mechanisms underlie this fundamental cognitive ability? Perceptual learning is an ideal model system for studying these questions. Performance in perceptual tasks improves with practice, long past the putative "critical period" in development. However, the improvement tends to be stimulus- and task-specific, which suggests the involvement of low-level sensory areas. This talk discusses the mechanisms of perceptual learning from an interdisciplinary perspective. I present data from behavioral experiments involving a novel non-stationary context manipulation. Orientation discrimination of Gabor targets embedded in filtered visual noise improved over the course of several days of practice, both with and without feedback. A neural network model provides an existence proof that incremental Hebbian reweighting can account quantitatively for the complex pattern of learning in this non-stationary training protocol. The model takes grayscale images as inputs, produces binary responses as outputs, and improves its discrimination accuracy incrementally with practice with no need for external feedback. Its performance is thus directly comparable to the human data. Learning occurs only in the read-out connections to a decision unit; the stimulus representations never change. This very simple, local learning rule converges to an approximately optimal classification policy. The broader implications of this statistical learning scheme are discussed. Reprints, data sets, and Matlab software are available at alexpetrov.com. _____
Add to calendar
Share|