Fig. 97

The model architecture (A) used to simulate the water maze task indicating the notion of a direct and an indirect pathway. The value function (B) developed after training the agent for 10 trials, the value peaks near to the platform. The escape latency (C) through trials shows that the agent has learnt the task with increased hippocampal dependence in the earlier stages and cortical dependence in the later stages of learning. The spectrogram (D) of the activity of CA3 as the function of time shows desynchronization while active exploration of the maze (0–45 s) and synchronized activity upon reaching the platform (45–65 s)