Estimating a Separable Random Field from Binary Observations

Summary

Date: 
November 22, 2016 - 12:00pm - 1:00pm
Location: 
Northwest 243
About the Speaker
Name: 
Demba Ba
Speaker Title: 
Asst. Professor of Electrical Engineering and Bioengineering
Speaker Affiliation: 
Harvard SEAS

Abstract:
Stochastic processes with dynamics at multiple time scales are pervasive in science and engineering. The spiking dynamics of neurons can exhibit variability both within a given trial (fast time scale) and across trials (fast time scale). The behavioral training of animals involves dynamics within the confine of a session (fast time scale) and across sessions (slow time scale). State-of-the-art methods for analyzing these data neglect their inherent two-dimensional nature by aggregating across one of the time scales. I propose a separable two-dimensional (2D) random field (RF) model of multiscale stochastic processes whose dynamics a different scales obey a certain separability condition. This model obviates the need to aggregate data across time scales and can characterize intricate dynamics present in the data. I demonstrate this model on data from two experiments. The first data set consists of neural spiking activity from neurons in the anterior cingulate cortex (ACC). The model is able to determine both the trial and the time within a trial when ACC neurons learn a fear response. The second data set comprises 14 female macaque monkeys (6 young) performing a reversal learning task in the form of a modified Wisconsin Card Sort Task. I demonstrate that, as a group, the old monkeys perceive the task as more difficult and are less flexible cognitively.