6e7d2da6d3953058db75714ac400b584-Reviews.html

Review for #1163 "Universal models for binary spike patterns using centered Dirichlet processes." The goal of this paper is to provide a more accurate method for modeling the distribution of binary spike patterns over a population of neurons. Essentially what the authors are trying to do is improve upon parametric models of the pattern distribution (such as a Bernoulli, cascaded logistic or Ising model) by allowing for deviations from the parametric model (or base model) if they are justified by the data. The involves postulating a Dirichlet process centered upon the base model and fitting the parameters of the base model and the concentration parameter of the Dirichlet process via gradient ascent (although I imagine other methods could be used for fitting). Intuitively this constitutes fitting a type of weighted average between the probability distribution of the base model and the pattern probabilities estimated by counting alone.