- Reinforcement Learning with TensorFlow
- Sayon Dutta
- 81字
- 2025-02-23 17:35:22
The policy model for optimality
Policy is defined as the model that guides the agent with action selection in different states. Policy is denoted as .
is basically the probability of a certain action given a particular state:

Thus, a policy map will provide the set of probabilities of different actions given a particular state. The policy along with the value function create a solution that helps in agent navigation as per the policy and the calculated value of the state.