Fig. 4
From: Cortical recruitment determines learning dynamics and strategy
A multiplicative reinforcement learning model reproduces modulations of learning speed by neuronal recruitment. a Schematics describing the auditory Go/NoGo discrimination model. b Mean±standard deviation of the difference of learning phase duration against difference in neural recruitment by the stimuli for the model initialized with 15 sets of initial conditions obtained when fitting learning curves from a previous study25. The experimental observations of Fig. 3d are superimposed in red. c Mean±standard errors for the learning and delay phases obtained with the model for a twofold difference in cortical recruitment between the two stimuli. A longer learning phase (Kolmogorov–Smirnov test, p = 8 × 10−7 indicated as ***, n = 15 initial conditions per group) but not delay phase (Kolmogorov–Smirnov test, p = 0.13 indicated as ns, n = 15 initial conditions per group) is observed when S− recruits less activity (blue) as compared to when S− recruits more activity than S+ (orange). d Learning phase duration plotted against the value of the modeled initial synaptic weights. Same simulations and color code as in c. Significant correlations were observed only when S− recruits less activity (blue), and for wCe (ρ = −0.61, p = 0.015, n = 15) and wCi (ρ = −0.70, p = 0.0034, n = 15). Error bars represent standard errors (SEM). Source data are provided as a Source Data file
