Figure 2.

Model-free network architecture. Based on both current state and possible actions, the “matrix maint” determines what to maintain in PFC. Based on the stored information in PFC, “matrix out” determines the next chosen action via PFC out. PVLV (consisting of multiple biological systems) evaluates the actions (critic) and helps train the model. See text for in-depth description and functions of the various components of the network. Detailed network architecture is highly similar to the PBDM model discussed later.