Reviews: Neurons Equipped with Intrinsic Plasticity Learn Stimulus Intensity Statistics

Neurons with Intrinsic Plasticity Learn Stimulus Statistics In the paper, the author(s) build a generative product-poisson-gamma model to recognize images with wide dynamic range, by extending a previously used Poisson mixture model. An equivalent neural circuit model with unsupervised learning was also proposed to implement EM-like algorithms to find the model parameters in generative model. In the learning rule, the synaptic weights are changed according to a Hebb like rule with synaptic scaling, and intrinsic parameters lambda (Eq.7 in the paper) are updated depending on the net input. It is very interesting that the fixed points of the network updating parameters can have the same form with EM algorithm. However, I have the following questions about this paper: 1.