buyers = []
for ii, p in enumerate(generate_buyer_prices_paper(20)):
buyers.append(LinearBlackBoxBuyer('b' + str(ii), p, noisy=True))
agents = sellers + buyers
for ag in agents:
ag.new_game()
ROUNDS_PER_GAME = 10
market_env = MarketEnvironment(
sellers=sellers,
buyers=buyers,
max_time=10, ## not the same as rounds per game!!
matcher=RandomMatcher(reward_on_reference=True))
init_observation = market_env.reset()
round_avg = 0.
offer_avg = 0.
time_avg = 0.
records['demands'].append([])
records['rewards'].append([])
records['prices'].append(seller_agent.reservation_price)
for n_round in range(10):
init_observation = market_env.reset()
observations = {k.agent_id: None for k in agents}
done = {k.agent_id: False for k in agents}
res_prices = np.random.normal(200, 5, n_buyers)
names = ['Buyer ' + str(i) for i in range(1, n_buyers + 1)]
buyers = np.array([LinearSameSideBuyer(agent_id=names[i], reservation_price=res_prices[i]) for i in range(n_buyers)])
# For plotting
fig, ax = plt.subplots(figsize=(8, 8), tight_layout=True)
ax.set_xlim(95, 205)
# Loop over games
for g in range(n_game):
print("GAME", g, '=================================================================================================================')
# Define parameters of each round
max_time = 30
matcher = RandomMatcher(reward_on_reference=False)
# Create market environment
market_env = MarketEnvironment(sellers=sellers, buyers=buyers, max_time=max_time, matcher=matcher)
# HERE AGENTS LEARN AND ADJUST THEIR COEFS (for now the are constant)
for agent in sellers:
size_coefs = agent.determine_size_of_coefs(n_sellers=n_sellers)
agent.coefs = np.array([0.05, 0.95] + [0]*(size_coefs - 2))
for agent in buyers:
size_coefs = agent.determine_size_of_coefs(n_buyers=n_buyers)
agent.coefs = np.array([0.05, 0.95] + [0]*(size_coefs - 2))
# Reset agents' rewards and observations
for agent in sellers:
agent.reward = 0.0