Analyzing statistical dependencies in neural populations

摘要: Analyzing Statistical Dependencies in Neural Populations by Ilan N. Goodman Neurobiologists recently developed tools to record from large populations of neurons, and early results suggest that neurons interact encode information jointly. However, traditional statistical analysis techniques are inadequate elucidate these interactions. This thesis develops two multivariate dependence measures that, unlike measures, encompass all high-order non-linear These decompose the contributions distinct subpopulations total dependence. Applying recordings crayfish visual system, I show neural exhibit complex dependencies vary with stimulus. Using Fisher analyze effectiveness population codes, optimal rate coding requires negatively dependent responses. Since positive through overlapping stimulus attributes is an inherent characteristic many systems, such can only achieve code cooperating.