|
Abstract:
|
It is well known that brain areas receive, process, and transmit information via sequences of sudden, stereotyped electrical impulses, called action potentials or spikes, and point process modeling and estimation algorithms have become pervasive in the analysis of such data. Recent technological developments have led to a massive increase in the size and dimensionality of neural datasets and focused researchers on models that can capture the structure of activity from of populations simultaneously recorded neurons. This has spurred the development of new modeling and estimation methods based on the theory of marked point processes.
In this talk, I will present a case study focused on understanding how mental exploration drives learning, which highlights a number of statistical problems that can be addressed under a marked point process modeling framework. I will discuss adaptive models that can capture nonstationarities in the data, goodness-of-fit methods that allow for model assessment and refinement, and state-space estimation methods that allow us to decode signals directly from the observed neural activity.
|
