a = Actor(LAMBDA, ALPHA_ACTOR, GAMMA, EPSILON)
c = create_critic(LAMBDA, ALPHA_CRITIC, GAMMA, USE_TABLE, state_size,
LAYERS, ADAPT_END_LAYERS, UPDATE_WHOLE_SEQUENCE)
if SCENARIO in SCENARIO_DESCRIPTIONS:
print(
f'\nScenario: {SCENARIO}\nDescription: {SCENARIO_DESCRIPTIONS[SCENARIO]}\n'
)
else:
print(
f'\nUndefined scenario\n\nRunning scenario: {0}\nDescription: {SCENARIO_DESCRIPTIONS[0]}\n'
)
print('Training')
results = run_ai(BOARD_TYPE, BOARD_SIZE, EPOCHS, a, c, EMPTY_NODES_POS)
a.epsilon = 0
if ITERATIONS_TEST:
print('\nTesting policy with greedy behaviour')
results += run_ai(BOARD_TYPE, BOARD_SIZE, ITERATIONS_TEST, a, c,
EMPTY_NODES_POS, False, False, False)
if ITERATIONS_SHOW_END_STATE:
print('\nTesting policy and showing only last move')
results += run_ai(BOARD_TYPE, BOARD_SIZE, ITERATIONS_SHOW_END_STATE, a,
c, EMPTY_NODES_POS, False, False, True, DEBUG,
DELAY_END_STATE)
if ITERATIONS_SHOW_ALL_ACTIONS:
print('\nTesting policy and showing every move')
results += run_ai(BOARD_TYPE, BOARD_SIZE, ITERATIONS_SHOW_ALL_ACTIONS,
