#QLearning_best['episode'] = episode
#QLearning_best['learning_rate'] = episode
#QLearning_best['best_value'] = episode
elif agent == sarsaAgent:
SARSA_stats['episode'].append(episode)
SARSA_stats['learning_rate'].append(step_size)
SARSA_stats['best_value'].append(episode_reward)
else:
expectedSARSA_stats['episode'].append(episode)
expectedSARSA_stats['learning_rate'].append(step_size)
expectedSARSA_stats['best_value'].append(episode_reward)
# EPSILON DECAY for exploratory
if params.END_EPSILON_DECAYING >= episode >= params.START_EPSILON_DECAYING:
params.set_epsilon(params.epsilon_decay_value)
# show some STATS
ep_rewards.append(episode_reward)
# Append the sum of reward at the end of the episode
#totalReward[type(agent).__name__].append(episode_reward)
all_reward_sums[(agent, step_size)].append(episode_reward)
if not episode % params.STATS_EVERY:
average_reward = sum(ep_rewards[-params.STATS_EVERY:]) / len(
ep_rewards[-params.STATS_EVERY:])
aggr_ep_rewards['ep'].append(episode)
aggr_ep_rewards['avg'].append(average_reward)
aggr_ep_rewards['max'].append(
max(ep_rewards[-params.STATS_EVERY:]))
SARSA_stats['episode'].append(episode)
SARSA_stats['learning_rate'].append(step_size)
SARSA_stats['best_value'].append(episode_reward)
SARSA_stats['discount'].append(discount)
SARSA_stats['epsilondecay'].append(epsilondecay)
else:
expectedSARSA_stats['episode'].append(episode)
expectedSARSA_stats['learning_rate'].append(step_size)
expectedSARSA_stats['best_value'].append(episode_reward)
expectedSARSA_stats['discount'].append(discount)
expectedSARSA_stats['epsilondecay'].append(epsilondecay)
# EPSILON DECAY for exploratory
if params.END_EPSILON_DECAYING >= episode >= params.START_EPSILON_DECAYING:
params.set_epsilon(epsilondecay)
# show some STATS
ep_rewards.append(episode_reward)
# Append the sum of reward at the end of the episode
#totalReward[type(agent).__name__].append(episode_reward)
#all_reward_sums[(agent, step_size)].append(episode_reward)
if not episode % params.STATS_EVERY:
average_reward = sum(ep_rewards[-params.STATS_EVERY:])/len(ep_rewards[-params.STATS_EVERY:])
aggr_ep_rewards['ep'].append(episode)
aggr_ep_rewards['avg'].append(average_reward)
aggr_ep_rewards['max'].append(max(ep_rewards[-params.STATS_EVERY:]))
aggr_ep_rewards['min'].append(min(ep_rewards[-params.STATS_EVERY:]))
#print(f'Agent: {type(agent).__name__}, Learning Rate: {params.LEARNING_RATE}, discount: {params.DISCOUNT:>2.2f}, episode: {episode:>5d}, average reward: {average_reward:>4.1f},min reward: {min(ep_rewards[-params.STATS_EVERY:])}, max reward: {max(ep_rewards[-params.STATS_EVERY:])},current epsilon: {params.EPSILON:>1.3f},epsilon_decay_value: {params.epsilon_decay_value:>1.5f}')
