Date: Wednesday, 05.07.23 17:00 CET
Location: Building S1|03 Room 223
Abstract:
In this talk I will discuss recent work from our group, in which we are using analysis of dynamics to generate insight into the neural computations underlying cognition. I will focus on two questions. First, how does synaptic pruning in cortex during adolescence, in which up to 40% of excitatory synapses are eliminated, affect cognition and neural dynamics. I will first discuss work in which we examined this question from a purely theoretical perspective, by pruning recurrent neural networks while they learn a working memory task and showing that pruned networks are more resistant to distractors. Pruned networks also show deeper basins around dynamical attractors, which accounts for their resistance to distraction. I will then discuss analysis of a longitudinal EEG dataset collected during adolescence and show that task related neural dynamics undergo changes that are consistent with the pruned recurrent neural networks. In the final part of the talk, I will describe work on the second question in which we used collected high-channel count multielectrode recordings in primate prefrontal cortex. We examined neural dynamics related to decision processes and found that decisions made with more certainty lead to deeper attractor basins. Overall, we find that neural dynamics are a useful tool for bridging theory and empirical data and generating insights into the neural mechanisms underlying cognitive processes.
