The Pleasures and Pains of Spike-Based Coding in the Nervous System: The Need for New Statistical Methodologies

• Activity Number: 114
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Theodore W. Berger*+
• Companies: University of Southern California
• Keywords: spike ; brain ; spatio-temporal pattern ; nonlinear ; dynamics

Abstract

Neurons communicate with each other primarily with all-or-none bioelectrical events known as action potentials. Action potentials (spikes) provide for a pulse-based coding scheme, because the shapes of spikes are nearly equivalent, with only inter-spike-intervals (ISI) varying systematically in the presence of different events. Thus, neural representations in the brain are based on temporal patterns, or the sequence of ISIs. Because neural representations never rely on a single neuron, and instead emerge from the activity of a population of neurons, neural representations are based, not on temporal patterns alone, but on spatio-temporal patterns. Although "neural processing" is an often-maligned term, its meaning here is straightforward: it is the consequence of changing an incoming spatio-temporal pattern into a different, outgoing spatio-temporal pattern. Because the response of any neuron depends not only on the most recent input event, but also on the preceding history of activity, neural processing involves nonlinear transformations that can reach high orders. This session will discuss new statistical methodologies demanded by analysis of systems with such properties.