Policy

conditional distribution over actions given the state: $P(A_t=\mathbf{a}|X_t=\mathbf{x})$