(A) Example clustering run for all neurons in the ephys dataset. First, unsupervised clustering is performed on the matrix of response reliabilities to four stimulus types. Colors on the t-SNE plot (used for visualization purposes only) represent different cluster labels, with one cluster highlighted in brown. Next, the neurons in each cluster are used to compute the mean reliability for each stimulus type. The mean reliabilities for all 27 clusters are shown in the heatmap, with the 25% responsiveness threshold indicated by the transition from blue to red. Each cluster is assigned to a class based on the stimulus types for which its mean reliability is above threshold. (B) Strip plots representing the fraction of neurons belonging to each class, averaged over 100 repeats of the Gaussian mixture model for ephys (N = 11,030 neurons), at a 25% threshold level. (C) Jensen–Shannon distance between the ephys and imaging class membership distributions, as a function of the ephys responsiveness threshold. The points plotted in panels B and E are circled. Heat maps show the mean reliabilities relative to threshold for all 27 clusters shown in panel A. (D) Strip plots representing the fraction of neurons belonging to each class, averaged over 100 repeats of the Gaussian mixture model for imaging (N = 20,611 neurons), at a 25% threshold level. (E) Same as B, but for the 40% threshold level that minimizes the difference between the imaging and ephys distributions.