Reinforcement Learning of Targeted Movement in a Spiking Neuronal Model of Motor Cortex

Figure 3

Performance across learning conditions and random wiring seeds.

A. End-of-trial errors under different learning conditions: no learning, reward-only, punisher-only, reward-and-punisher; for each: 5 target angles (0°, 35°, 75°, 105°, 135°)×5 random wirings×5 random babble noise inputs. B. Learning performance as a function of the percentage of EM cells that have low-convergence, defined as having fewer than five afferent inputs. The final error appears to be strongly correlated with the proportion of low-convergence cells (; Pearson’s , ).

doi: https://doi.org/10.1371/journal.pone.0047251.g003