Lunch is available beginning at 12 PM

Speaker to begin promptly at 12:30 PM

This talk will feature discussions from Jonathan Pillow and Helena Liu.

Abstract: Building on insights into behavioral learning discussed earlier, we will delve into the synaptic mechanisms underlying the brain's remarkable ability to learn and adapt. A central challenge is the 'temporal credit assignment' problem: how do neural circuits determine which specific states and connections contribute to future outcomes and subsequently adjust these for enhanced learning? Leveraging fresh insights from the Allen Institute’s data, we propose a novel learning rule for recurrent neural networks based on widespread cell-type-specific local neuromodulatory interactions, predicting their roles in synaptic credit assignment. Additionally, we will discuss our investigations into the generalization properties of biologically plausible learning systems, utilizing insights from recent deep learning theory regarding loss landscape curvature. We will also explore the influence of initial connectivity structure, especially its effective rank, on learning dynamics. Lastly, time permitting, I will briefly discuss our ongoing work on inferring learning rules from behavioral data.

Contributions to and/or sponsorship of any event does not constitute departmental or institutional endorsement of the specific program, speakers or views presented.