Workshop in Biostatistics presents "Neural Data Analysis— From Stochastic Process Modeling to Deep Learning"

Topic: 
Neural Data Analysis— From Stochastic Process Modeling to Deep Learning
Thursday, December 5, 2019 - 1:30pm
Venue: 
Medical School Office Building (MSOB), Rm x303
Speaker: 
Babak Shahbaba (Professor of Statistics and Computer Science at UC Irvine)
Abstract / Description: 

In this talk, I will discuss an ongoing project aimed at understanding the neural basis of complex behaviors and temporal organization of memories. More specifically, I will focus on a unique electrophysiological experiment designed to address fundamental and unresolved questions about hippocampal function. Our goal is to elucidate the neural mechanisms underlying the memory for sequences of events, a defining feature of episodic memory. To this end, we have used high-density electrophysiological techniques to record neural activity (spikes and local field potentials) in hippocampal region CA1 as rats perform an odor sequence memory task. Importantly, this nonspatial approach allows us to determine whether spatial coding properties (thought to be fundamental to hippocampal memory function) extend to the nonspatial domain. To answer this question, we have developed a set of flexible inferential methods based on Gaussian process models for detecting neural patterns and a set of powerful predictive models based on deep learning algorithms for neural decoding. Our findings could lead to unprecedented insight into the neural mechanisms underlying memory impairments.

 

Suggested Readings:
● https://www.jneurosci.org/content/36/5/1547 (Journal of Neuroscience)
● https://arxiv.org/pdf/1711.02869.pdf (to appear in Bayesian Analysis)
● https://arxiv.org/pdf/1905.10413.pdf (to appear in NeurIPS 2019)