Probabilistic Spike Train Inference

Recent advances in neuroscience have enabled researchers to measure the activities of large numbers of neurons simultaneously in behaving animals. Scientists have access to the fluorescence of each of the neurons that provides a first-order approximation to the neural activity over time. Determining a neuron’s exact spike times from this fluorescence trace is an important task under the realm of computational neuroscience. We propose a novel Bayesian approach based on a mixture of half-non-local prior densities and point masses for this task, adopt a stochastic search based approach to report highest posterior probability arrangement of spikes. Our proposals, besides providing automatic uncertainty quantification associated with the spike train inference, lead to substantial improvements over existing proposals based on L
1
regularization, and enjoys comparable estimation accuracy to the state-of-the-art L
0
proposal, in simulations and on recent calcium imaging data sets.