def __init__(self, bidder_id, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc):
"""
:param bidder_id: Integer. A unique identifier for this given agent.
:param num_rounds: Integer. The number of rounds the auction this bidder is participating in will run for.
:param num_bidders: Integer. The total number of bidders in the auction this bidder is participating in.
:param possible_types: List. A list of all possible types the bidder can take. Types are arranged in
increasing order.
:param type_dist: List. Probabilities corresponding to each entry in possible_types.
:param type_dist_disc: Boolean. True if type_dist is describing a discrete distribution.
"""
# Init will call make_state_space, but we need some more information before going through this process.
MDPBidderUAI.__init__(self, bidder_id, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
self.num_price_samples = len(self.action_space) # Not used
self.num_prices_for_state = len(self.action_space) - 1
state_price_delta = float(max(possible_types) - min(possible_types)) / self.num_prices_for_state
self.prices_in_state = [i * state_price_delta for i in range(self.num_prices_for_state)]
self.digit_precision = 2 # Learn values to this precision
self.prices_in_state = set()
self.state_space = set()
self.terminal_states = set()
self.action_space = set()
self.use_given_pi = False
self.given_pi = {}
if type_dist_disc:
type_dist = [1.0 / len(possible_types)] * len(possible_types)
else:
type_dist = [1.0] * len(possible_types)
num_mc = 50000
bidders = [KatzmanBidder(i, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
for i in range(num_bidders)]
# bidders = [MenezesMonteiroBidder(i, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
# for i in range(num_bidders)]
# bidders = [SimpleBidder(i, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
# for i in range(num_bidders)]
# bidders = [WeberBidder(i, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
# for i in range(num_bidders)]
learner = MDPBidderUAI(num_bidders, num_rounds, num_bidders, possible_types, type_dist, type_dist_disc)
learner.learn_auction_parameters(bidders, num_mc)
# Plot what the bidder has learned
learner.valuations = [.2, .1]
learner.calc_expected_rewards()
learner.solve_mdp()
plot_exp_payments(learner)
plot_transition(learner)
plot_prob_winning_and_transition(learner)
plot_Q_values(learner)
plot_price_pdf(learner)
print(learner.place_bid(1))
# print(learner.place_bid(2))
# Compare learner to other agents
