Figure 2.

Subunit structures of the two RGC models revealed by the CNN and CRNN models

(A) RFs of the two RGCs.

(B) Overlaid RF subunits of the two RGCs.

(C) Comparison of the measured spatial and temporal STAs of the two RGCs with those predicted by the CNN and CRNN.

(D) The RF subunits of model I with the convolutional filters learned in the CNN and CRNN models, with a kernel size of 6.

(E) Spatial autocorrelation versus temporal regularity of the model and convolutional filters in the models.

(F and G) Similar to (D) and (E), respectively, but for model II, in which the first subunit is changed, followed by altering all eight RF subunits in the models. RF, receptive field; STA, spike-triggered average.