750aa623cb885d273d52c85da3cfc4a958ec836f,ch05/02_frozenlake_q_learning.py,,value_iteration,#Any#Any#,52

Before Change

    for state in range(agent.env.observation_space.n):
        for action in range(agent.env.action_space.n):
            action_value = 0.0
            target_counts = agent.transits[TransitKey(state, action)]
            total = sum(target_counts.values())
            for tgt_state, count in target_counts.items():
                reward = agent.rewards[RewardKey(state, action, tgt_state)]
                best_action = agent.select_action(tgt_state, values)
                action_value += (count / total) * (reward + GAMMA * values[ValueKey(tgt_state, best_action)])
            values[ValueKey(state, action)] = action_value

After Change

    for state in range(agent.env.observation_space.n):
        for action in range(agent.env.action_space.n):
            action_value = 0.0
            target_counts = agent.transits[(state, action)]
            total = sum(target_counts.values())
            for tgt_state, count in target_counts.items():
                reward = agent.rewards[(state, action, tgt_state)]
                best_action = agent.select_action(tgt_state, values)
                action_value += (count / total) * (reward + GAMMA * values[(tgt_state, best_action)])
            values[(state, action)] = action_value

Instances