Fig. 7
Comparisons of performance between minimal OLE, Direct Regression and ANN using arm movement datasets. a, i Distribution of preferred directions for the selected 33 units for Monkey N and for 19 units of Monkey C. The red line represents the mean of each distribution. c, k Reconstructed trajectories using minimal OLE for monkeys N and C. d, l Speed profiles using minimal OLE for each monkey. e, m Reconstructed trajectories using Direct Regression. f, n Speed profiles using Direct Regression. g, o Reconstructed trajectories using the ANN for each monkey. h, p ANN Speed profiles. b, j Velocity correlation coefficients between decoded and actual data for all trajectories of monkeys N and C. The correlation coefficients from the ANN decoder were significantly higher than those of the OLE. (p < 0.001, t-test, Bonferroni). Error bars denote the 95% confidence interval over all trials (n = 69 × 10 trials per target for Monkey N and 47 × 10 trials for Monkey C from the 10 repetitions of cross-validation decoding analysis with random data partitions)
