Fig. 3: Networks with larger repertoires can adapt to perturbations more easily.
From: De novo motor learning creates structure in neural activity that shapes adaptation
a Motor output of networks trained on different repertories (legend) following adaptation to a counterclockwise VR perturbation. b Loss during adaptation, calculated as the mean-squared error between the network output and target positions. Line, shaded surfaces, smoothed mean, and 95% confidence interval across networks of different seeds. c Decay constants for exponential curves fitted to the loss curves in panel (b). Circles and error bars, means and 95% confidence intervals with bootstrapping. *** denotes 0.001, ** 0.01, * 0.05 for two-sided paired t-tests. Data includes n = 10 networks from ten random seeds for each repertoire.
