Fig. 7: Recurrence enables better generalisability to stochastic environments.
From: Hippocampus supports multi-task reinforcement learning under partial observability
a Change in state-occupancy (with probabilistic cues - without probabilistic cues) for the hcDRQN agent across different degrees of cue removal. b hcDRQN outperforms hcDQN across different degrees of probabilistic cues. c Performance of hcDRQN agents trained with probabilistic cues compared to without. The hcDRQN was trained 80% cue probability. d Change in state-occupancy of hcDRQN agents trained with probabilistic cues. Data are presented as mean values  ± SEM over 5 different initial conditions. Source data are provided as a Source Data file.
