An overview of the model.

<p>(A) The neural circuit receives normalized inputs conveyed by excitatory synapses to a processing layer (large figure). The activity of the processing neurons is determined by the received inputs and internal dynamics mediated by lateral interactions. Inset: Two forms of weight scaling. The red curve shows conventional linear scaling, the green curve logarithmic scaling for values larger one. (B) Inputs to the circuit are modeled using a mixture model with normalized generative fields and Poisson noise. (C) Example normalized fields, with different values of the normalization constant . (D) Illustration how inputs with different contrast levels are normalized (background set to 1).