Skip to main content
. 2022 Feb 14;16:760864. doi: 10.3389/fnsys.2022.760864

FIGURE 6.

FIGURE 6

Activity of RNNs during working memory tasks. (A) Mean firing rate of RNN units responsive to the ODR task for the location corresponding to the best delay period activity of each unit in fully trained networks. Solid line indicates PSTH of correct trials, dashed line indicates PSTH of error trials. N = 2,704 units for correct trials, 244 units for error trials. Inset on top of figure indicates schematically the sequence of events in the task; location of stimulus used to construct the mean firing rate plot differed for each unit. (B) As in (A) for an RNN network trained with a variable delay period. N = 5,039 units for correct trials, 144 units for error trials. (C) Mean firing rate of RNN units responsive to the ODRD task. Solid line indicates PSTH of correct trials when cue is in the best location of each unit; dashed line indicates PSTH of error trials. N = 2,951 units for correct trials, 208 units for error trials.