%0 Software %A Rostami, Vahid %E Mana, PierGianLuca %E Helias, Moritz %D 2016 %T Pairwise maximum-entropy models: bimodality, bistability, non-ergodicityproblems, and their elimination via inhibition %X The pairwise maximum-entropy model [1,2], applied to experimental neuronal data of populations of 200 andmore neurons, is very likely to give a bimodal probability distribution for the population-averaged activity. Wehave provided evidence for this claim, starting from an experimental dataset and then looking at summarizeddata from the literature. The first mode is the one observed in the data. The second mode (unobserved)can appear at very high activities (even 90% of the population simultaneously active) and its height increaseswith population size. This bimodality has several undesirable consequences:1.The presence of two modes is unrealistic in view of observed neuronal activity.2.The prediction of a high-activity mode is unrealistic on neurobiological grounds.3.Boltzmann learning becomes non-ergodic, hence the pairwise model found by this method is not themaximum entropy distribution; similarly, solving the inverse problem by common variants of mean-fieldapproximations has the same problem.4.The Glauber dynamics associated with the model is either unrealistically bistable, or does not reflect thedistribution of the pairwise model.