def qlearn_learn(nplayers, num_card_values, num_cards, agent, learn_trials, test_trials, featureExtractor=qlearn.bsFeatureExtractor, explorationProb=0.2, maxIters=1000, dishonesty_list=None, confidence_list=None, learn_list=None, verbose=False):
print "Running qlearning as agent", agent, "."
game = play_game.BSGame(nplayers, [num_cards for _ in range(num_card_values)], agent, verbose=0)
if not dishonesty_list:
dishonesty_list = [0.1] * nplayers
if not confidence_list:
confidence_list = [1] * nplayers
if not learn_list:
learn_list = [False] * nplayers
print "Players have dishonesty", dishonesty_list
# print "Players have confidence", confidence_list
apolicies = []
for t in range(nplayers):
apolicies.append(policy.DishonestPolicy(game, dishonesty_list[t], confidence=confidence_list[t], learn=learn_list[t]).decision)
game.setPolicies(apolicies)
qlearning = qlearn.QLearningAlgorithm(game.actions, game.discount(), featureExtractor, explorationProb=explorationProb)
print "Learning..."
simulators.rlsimulate(game, qlearning, numTrials=learn_trials, verbose=False, maxIterations = maxIters)
qlearning.explorationProb = 0
game.resetWins()
print "Learning complete. Now simulating tests..."
simulators.rlsimulate(game, qlearning, numTrials=test_trials, verbose=verbose, maxIterations = maxIters)
print "Wins observed:", game.wins
print "Agent in position", agent, "has a win rate of", str(float(game.wins[agent])/sum(game.wins)), '\n'
return qlearning
def qlearn_test(nplayers, num_card_values, num_cards, agent, trials, qlearning, dishonesty_list=None, confidence_list=None, learn_list=None, verbose=False):
print "Testing qlearning as agent", agent, "."
game = play_game.BSGame(nplayers, [num_cards for _ in range(num_card_values)], agent, verbose=0)
if not dishonesty_list:
dishonesty_list = [0.1] * nplayers
if not confidence_list:
confidence_list = [1] * nplayers
if not learn_list:
learn_list = [False] * nplayers
print "Players have dishonesty", dishonesty_list
# print "Players have confidence", confidence_list
apolicies = []
for t in range(nplayers):
apolicies.append(policy.DishonestPolicy(game, dishonesty_list[t], confidence=confidence_list[t], learn=learn_list[t]).decision)
game.setPolicies(apolicies)
qlearning.explorationProb = 0
print "Simulating..."
simulators.rlsimulate(game, qlearning, numTrials=trials, verbose=verbose)
print "Wins observed:", game.wins
print "Agent in position", agent, "has a win rate of", str(float(game.wins[agent])/sum(game.wins)), '\n'