Fig. 7: SNNs constrained to the orientation space do not outperform the mouse.
From: Discretized representations in V1 predict suboptimal orientation discrimination
a Schematic representation of the shallow neural network trained to classify the orientations of the tuning block stimuli. b Performance of the SNNs at classifying the tuning curve stimuli. c D’ of the same SNNs as in k. when classifying the Go and NoGo stimuli. In brown, the best possible D’ for each SNN computed from a ROC analysis (see Fig. S15). The violin plots show the distribution of values obtained from repeating the classification 1000 times. In blue is reproduced the animal’s performance as in Fig. 1 for comparison.
