Greg Jensen, Fabian Munoz, Vincent P Ferrera
The electrophysiological study of learning is hampered by modern procedures for estimating firing rates: Such procedures usually require large datasets, and also require that included trials be functionally identical. Unless a method can track the real-time dynamics of how firing rates evolve, learning can only be examined in the past tense. We propose a quantitative procedure, called ARRIS, that can uncover trial-by-trial firing dynamics. ARRIS provides reliable estimates of firing rates based on small samples using the reversible-jump Markov chain Monte Carlo algorithm. Using weighted interpolation, ARRIS can also provide estimates that evolve over time. As a result, both real-time estimates of changing activity, and of task-dependent tuning, can be obtained during the initial stages of learning.