INNER CODE UNIT · Python

policy

rlcode/reinforcement-learning · 1-grid-world/2-value_iteration.py:70

        policy = [[[0.0] * 4 for _ in range(env.width)] for _ in range(env.height)]
        for state in env.get_all_states():
            x, y = state
            actions = value_iteration.get_action(state)
            if not actions:
                continue
            prob = 1.0 / len(actions)
            for a in actions:
                policy[x][y][a] = prob
        display.show_arrows(policy)

    def on_move():
        display.move_along_policy(value_iteration.get_action)

    def on_clear():
        value_iteration.__init__(env)
        display.clear()
        display.agent_pos = [0, 0]

View source record →