Figure 3.

Encoding and decoding filters for real and model neurons. A, Encoding filters of three auditory neurons (left) and their one-dimensional, frequency-only and normalized approximations: the SRFs (right). B, Decoding filters of the same three auditory neurons (left) and their one-dimensional, frequency-only, and normalized approximations (right). They are much wider than the corresponding encoding filters in A. C, Inference model neurons, whose selectivities are given by the PFs in B, have narrow encoding filters (similar to real encoding filters). D, Decoding filters of the same model neurons are wider than the encoding filters in C and resemble the decoding filters of real neurons in B.