Fig. 8

Performance of the distributed model decoder across different spatial cues. a Confusion matrices showing the % of decoded location classifications of each stimulus location when the distributed decoder was trained with responses to low-pass (LPN) stimuli and tested with responses to the high-pass (HPN) stimuli from the same units that had significant spatial information in at least one bin width (top left), decoder trained on HPN responses and tested on LPN responses (top right), trained and tested on LPN (lower left) and trained and tested on HPN (lower right). b Root-mean squared error (RMSE) of the decoder (coloured symbols) was lower than when the cell identities were shuffled (grey) in each condition tested. The bars show the mean ± s.d. across 1000 iterations for observed and cell-identity shuffled decoding. P-values indicate the probability that the difference between the observed and shuffled mean RMSE is greater than the difference obtained by chance (estimated by permuting the labels between real and shuffled decoder scores (1000 iterations) and calculating the difference in decoding, see Methods). c Mean ± s.d. % correct of the decoder (filled bars) compared with % correct when cell identity was shuffled (grey bars). P-values indicate the probability that the difference between the actual and shuffled mean % correct would be observed by chance. The grey dotted line indicates chance performance (1/6 chance of correct classification)