# environment.renderer = renderer
# renderer.start()
# experiment.runEpisodes(4)
# run experiment
for i in range(5000):
experiment.runEpisodes(10)
agent.learn()
agent.history.keepBest(50)
# agent.forget()
valdata = experiment.evaluateEpisodes(10, visualize=False)
if renderer.isAlive():
environment.renderer = renderer
experiment.evaluateEpisodes(1, visualize=False)
environment.renderer = None
print i
print "mean return", mean([sum(v.rewards) for v in valdata])
if mean([sum(v.rewards) for v in valdata]) > 1.75*maxSteps:
if not renderer.isAlive():
renderer.start()
pass
print "avg. episode length", mean([len(v) for v in valdata])
print "exploration variance", explorer.sigma
print
