An Implementation of the Extended Poisson Model for Distributional Population Codes

We examine the Extended Poisson Model proposed by [2] for encoding and decoding probability distributions in populations of neurons. We derive the equations for two MAP Estimation methods: ExpectationMaximization and Softmax, to reconstruct an input distribution based on the observed firing activity of the neurons. We analyze the performance of these decoding methods in extracting information regarding multiple values and uncertainty in the value of the stimulus variable encoded, and demonstrate that it is robust to noise. We present the results which show that our implementation is able to reproduce the accuracy obtained in the original paper. In addition, we also examine the effect of smoothness priors and weighting factors used in the model, in the convergence rate of the algorithms and quality of reconstructed distribution.