FIGURE 8.
Texture tuning in the RFAMpa hidden layers (H1, H2) and V1, V2. Texture tuning quantified the average difference in response to examples from a given MSCT class and the random class. (A) Mean texture tuning for the six MSCT classes for the units in H1 of the RFAMpa. (B) Mean texture tuning of V1 neurons, grouped by cortical layer (legend). (C) Same as (A), now for H2 (D) Same as (B), now for V2. Error bars indicate standard error over units/neurons. Texture tuning was particularly strong for the visually salient MSCT (white triangle, black triangle, even, and odd) and stronger for H2 than H1. This qualitatively matches the MSCT selectivity properties of V1 and V2 neurons.