Fig. 3

Comparison of the subspace to a shuffled subspace. a State space showing the target locations in Delay 2 (dark colored dots) were shuffled relative to those in Delay 1 (light colored dots) before the optimization was performed in the full space. b Heat map showing the cross-temporal population-decoding performance in the full space with the target location in Delay 2 shuffled relative to those in Delay 1. c Delay 1 (plotted using crosses) and shuffled Delay 2 (plotted using open circles) responses after projection into the top 2 dimensions of the subspace. Points for different target locations are color-coded according to the color scheme shown in the top left. d Heat map showing the cross-temporal population-decoding performance in the subspace with the target locations in Delay 2 shuffled relative to those in Delay 1. e The difference in decoding performance when a decoder that was trained on the Delay 1 responses shown in c was tested on Delay 1 (LP₁₁) or Delay 2 (LP₁₂) responses are shown in the gray box-plot labelled LP₁₁ - LP₁₂, and vice versa (labelled LP₂₂ - LP₂₁). f The shift in cluster centers from Delay 1 to Delay 2 in the subspace, normalized to the average intra-delay distance, is shown for the true data (black box-plot) and the shuffled data (gray box-plot). The boxplot in the middle illustrates the intra-delay shifts for comparison. g The magnitude of the vector obtained by averaging the trajectory vectors from Delay 1 to Delay 2 for each target location is shown for the true data (black box-plot) and the shuffled data (gray box-plot). The bounds in the boxplots are defined by the 25th and 75th percentile of the distribution. The red line represents the median and the whiskers represent the 2.5th and 97.5th percentile of the distribution