Location: 46-3037 (NOTE: THIS IS NOT THE USUAL ROOM)

Zoom: https://mit.zoom.us/j/97203999531

Speaker: Antoine De Comite

Affiliation: Seethapani Lab

Title: A task-general learning rule for efficient motor learning

Abstract: Whether learning to ride a bike or to throw a baseball, human motor learning proceeds in episodic bouts of variable duration. When learning new movements, we spontaneously decide whether to swiftly discard a deteriorating bout of movement or to continue exploiting a high-value one. Here, we introduced a learning rule called ‘viability aware termination’ which decides, based on the quality of the current bout of movement, whether it is worth continuing. We show through simulations that this learning rule captures human motor behavior during learning, the characteristics of the episodic structure of learning, and yields more sample and energy efficient control strategies.