Analysis of Neuronal Spike Trains using Prototype Point Processes

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

The investigator develops tools for the summary and description of datasets involving catalogs of multiple realizations of point processes, using non-parametric techniques involving point process prototypes. Particular attention is paid to the case where each realization is a list of points in space and time, and a major characteristic of this proposed research involves the extension of prototypes and analyses based on metrics such as spike-time distance from one-dimensional point processes to the case of repeated realizations of point processes in higher dimensions. For such datasets, prototypes represent a useful summary of the behavior of the typical realization of the point process, and as such can be used to compare realizations from distinct classes of point processes. The methods investigated are entirely non-parametric, relying on minimal assumptons about the point processes being studied. The investigators apply their techniques to the neuroscience problem of summarizing and describing patterns in neuronal cell firings in conditioned and unconditioned subjects responding to various stimuli and to the description of typical spatial-temporal incidence of wildfires within seasons or years, as well as the characterization of other point process datasets such as earthquake catalogs. The methods developed in this proposal are useful for detecting and summarizing patterns in data from a wide variety of applications where one records repeated observations of phenomena that seem to occur at random times and locations. Examples include the firings of neurons within the brain, global earthquakes and their aftershocks, and the locations and times of wildfire ignitions in Southern California. Prototype methods are incredibly useful for the description of such catalogs and, in the context of earthquakes and wildfires, can aid urban planners and emergency response personnel, structural engineers and those interested in insurance in describing what to expect in a typical realization. In the neuroscience setting, this project contributes to the understanding of the brain and its properties, as the methods explored here help to characterize the typical neuronal firing patterns among subjects receiving different stimuli or different forms of conditioning.