11/3/2009
Department Seminar
11:45 AM
Whitaker 100
"Algorithms for learning in human motor adaptation"
Maurice Smith
Harvard University
Abstract: If we really understood the algorithms that our nervous systems used to learn motor skills, we might be able to this knowledge to teach and rehabilitate these skills more efficiently. This talk will present recent experimental and computational work which aim at this goal. I’ll focus on several key elements the of the algorithms that the human brain uses when learning simple motor skills, and attempt to provide a computational framework for an integrated understanding. These elements include: (1) The time-scales for the formation and decay of motor memories, and what these timescales tell us about the interactions between the adaptive processes underlying them. (2) How coupled representations of motion state (position and velocity) in the neural coding of motion determine (a) the fine temporal structure of motor adaptation (b) the ability of learn arbitrary patterns of dynamics and (c) the patterns of the motor variability during repeated performance. (3) How the coordinate system for the assignment of credit and blame for motor errors during learning controls the rate of learning. Recent studies that leverage the basic knowledge from this work to design learning paradigms which optimize learning and retention rates for motor skill acquisition will also be presented.
Biomedical Engineering
Back to Calendar