Fig A. Selecting stable trials of spontaneous activity. An example of the spontaneous LFP selection criteria for a single recording session. The solid blue line is the trial-average of the signals during each ISI. The vertical green line is stimuli o

Fig A. Selecting stable trials of spontaneous activity. An example of the spontaneous LFP selection criteria for a single recording session. The solid blue line is the trial-average of the signals during each ISI. The vertical green line is stimuli onset. The light blue shading is 2 st. dev. of the mean. The gray box is the stable time window chosen as the spontaneous activity window. Fig B. Decoding context fails with linear SVM. Average accuracy was computed over 50 sample linear SVMs. Trials were randomly reassigned to the training, validation, and testing sets for each sample classifier using an 80-10-10 split. Accuracy results for all recording sessions using the CNN classifier. The 50% quantile of accuracy is represented by a dot, with vertical bars reporting the 10% and 90% quantiles. Colors indicate the nucleus in which the recording contacts were located. Gray bars indicate the null distribution obtained from bootstrapping. Fig C. Decoding accuracy is a function of data av