A Mixed-Filter Algorithm for Dynamically Tracking Learning from Multiple Behavioral and Neurophysiological Measures

Abstract

Learning is a dynamic process generally defined as a change in behavior as a result of experience. Behavioral performance is commonly measured with continuous variables (reaction times) as well as binary variables (correct/incorrect task execution). When neural activity is recorded at the same time as behavioral measures, an important question is the extent to which neural correlates can be associated with the changes in behavior. Recent work has combined subsets of the three aforementioned modalities to understand learning. In this work, we develop an analysis of learning within a state-space framework of simultaneously recorded continuous and binary performance measures along with neural spiking activity modeled as a point process. This chapter illustrates our approach in the analysis of a simulated learning experiment, and an actual learning experiment, in which a monkey rapidly learns new associations within a single session.