def train(self):
MiscUtils.rm_hist()
for epi in range(20000):
s = self.sim.reset()
if self.best_score > Config.max_fitness():
break
while 1:
a = self.renet.choose_action(s)
s_, done, r = self.sim.step(a)
self.renet.store_transition(s, a, r, s_)
if self.renet.memory_counter > MEM_CAP:
self.renet.learn()
if done:
print('episode: {} score: {}'.format(
epi, self.sim.travel_range))
if self.sim.travel_range > self.best_score:
self.best_score = self.sim.travel_range
print('*' * 20)
print('New best score! score: {}'.format(
self.best_score))
print('*' * 20)
self.sim.save_gif()
if done:
break
s = s_
MiscUtils.finish_info()
def train(self):
def plot():
plt.cla()
if len(self.range_hist) > 2000:
self.range_hist.pop(0)
self.range_hist.append(self.sim.travel_range)
plt.plot(range(len(self.range_hist)),
self.range_hist,
linewidth=0.5)
plt.plot(range(len(self.range_hist)), self.range_hist, 'b^-')
plt.xlabel('Testing number')
plt.ylabel('Fitness')
plt.pause(0.01)
if self.sim.travel_range > Config.max_fitness():
plt.savefig('res/rl_statistics.png')
MiscUtils.rm_hist()
print('*' * 50)
print('Gathering experience...')
print('*' * 50)
for epi in range(20000):
s = self.sim.reset()
if self.best_score > Config.max_fitness():
break
while 1:
a = self.renet.choose_action(s)
s_, done, r = self.sim.step(a)
self.renet.store_transition(s, a, r, s_)
if self.renet.memory_counter > MEM_CAP:
self.renet.learn()
if done:
plot()
print('episode: {} score: {}'.format(
epi, self.sim.travel_range))
if self.sim.travel_range > self.best_score:
self.best_score = self.sim.travel_range
print('*' * 20)
print('New best score! score: {}'.format(
self.best_score))
print('*' * 20)
self.sim.save_gif()
if done:
break
s = s_
MiscUtils.finish_info()
def train(self):
MiscUtils.rm_hist()
local_dir = os.path.dirname(__file__)
config_path = os.path.join(local_dir, 'config-feedforward')
config = neat.Config(neat.DefaultGenome, neat.DefaultReproduction,
neat.DefaultSpeciesSet, neat.DefaultStagnation,
config_path)
p = neat.Population(config)
p.add_reporter(neat.StdOutReporter(True))
stats = neat.StatisticsReporter()
p.add_reporter(stats)
p.add_reporter(neat.Checkpointer(50))
winner = p.run(self.eval_genomes, 5000)
MiscUtils.finish_info()
def train(self):
MiscUtils.rm_hist()
print('*' * 50)
print('Starting RF Learning')
print('*' * 50)
for epi in range(2000000):
radar_data = self.sim.reset()
if self.best_score > self.max_fitness:
break
while True:
action = self.renet.choose_action(radar_data)
radar_data_, done, r = self.sim.step(action)
self.renet.store_transition(radar_data, action, r, radar_data_)
if self.renet.memory_counter > MEM_CAP:
self.renet.learn()
if done:
print('episode: {} score: {}'.format(
epi, self.sim.travel_range))
if self.sim.travel_range > self.best_score:
self.best_score = self.sim.travel_range
print('*' * 20)
print('New best score! score: {}'.format(
self.best_score))
print('*' * 20)
self.sim.save_gif()
if done:
break
radar_data = radar_data_