Problems at the Interface of Machine Learning and Neuroscience

The Spring 2022 SDS Seminar Series concludes on Friday, April 29 from 2:00 p.m. to 3:00 p.m. with Dr. John Lafferty (John C. Malone Professor of Statistics and Data Science at Yale University). This event is in-person, but a virtual option will be available as well. 

Title: Problems at the Interface of Machine Learning and Neuroscience

Abstract: A two-way channel exists between neuroscience and machine learning. In one direction, the processes and mechanisms of neural computation can serve as abstractions and sources of new mathematical frameworks for data analysis. In the other direction, computational modeling can be used to understand empirical findings in neuroscience, from the cellular level across species to the cognitive level in humans.  We present work on three problems at this interface.  First, we analyze a nonstandard algorithm for training neural networks that has been proposed as an alternative to the biologically implausible backpropagation algorithm. Second, we develop a computational model of collision detection in the compound eye, showing how optimization of inference tasks that are important for an animal’s perceptual goals can reveal and explain computational properties of specific sensory neurons. Third, we investigate how local patterns of excitation and inhibition may act as a computational mechanism underlying the organization of neuronal maps that has been observed in various brain regions across species. Finally, we comment on the potential for further alignment between machine learning and neuroscience in the future.