Figure 8.

Clustering of time-varying STRFs. (A) Temporal evolution of clusters of bootstrapped local STRFs for the latter two A1 example units in Figure 5B. The majority of clusters appeared only once during the whole stimulus sequence. The temporal extent of the clusters indicates that consecutive local STRFs describe very similar states. On a global scale, however, fluctuations appear at random and are not systematic. Clusters have been sorted by total size in descending order. Color encodes the number of bootstrap STRF estimates in the different clusters per time step (white: 0, black: 10). (B) Summary of cluster statistics across all clusters for 30 A1 units. Each histogram relates the number of occurrence of clusters to the total temporal extent of the clusters. For each unit, the total time the time-varying STRF was in a specific state (cluster) was estimated by counting all time steps in which more than half of the bootstrap estimates were assigned to the corresponding cluster. Note that we did not include clusters with the static STRF into the analysis (cluster 1 in both examples in A). (C) Same analysis as in b for 30 IC units. In IC, only a few clusters were identifiable, and virtually all clusters did not occur more than once per stimulus sequence. Thus, in both regions a time-varying STRF analysis does not reveal repeated patterns or long-term changes in STRF structure.