Machine Learning at the Flatiron Institute Seminar: Grace Lindsay

Title: Analyzing artificial neural networks to understand the brain

Abstract: In the first part of this talk, I will present work showing that artificial neural networks with recurrent connections can replicate broad behavioral patterns associated with dynamic visual object recognition in humans. An analysis of these networks shows that different types of recurrence use different strategies to solve the object recognition problem. The analysis of these networks, however, introduces another question: are the tools of neuroscience suitable for understanding complex distributed information processing systems? In the second part of this talk, I will discuss a research plan for testing a wide range of analysis tools frequently applied to neural data on artificial neural networks. I will present the motivation for this approach as well as the form the results could take and how this would benefit neuroscience and the field of interpretable AI.