Andrea Liu – University of Pennsylvania
Wednesday, September 27th, 10:45AM EDT (Zoom meeting starts at 10:30 EDT)
Learning about learning with physical networks
Unlike artificial neural networks, the brain does not require a computer in order to learn. Our experimental collaborators have developed laboratory electrical circuits that can learn to perform machine learning tasks without using a processor or external memory. They are far simpler than the brain but share the property that each component in the system does not need to know the state of every other component in order to learn. Our systems learn by themselves—the components update by identical local rules that lead to the emergence of learning as a collective phenomenon. I will describe what we have learned about the learning process in these systems