def playArm(self, env, t):
assert isinstance(env, UnimodalEnvironment
), "not unimodal environment for unimodal policy"
nb_arms = env.nb_arms
if t < nb_arms:
arm_t = env.get_arm_idx(t)
leader_t = None
else:
list_arm_mu_hat = [(arm, arm.mu_hat) for arm in env.list_of_arms]
leader_t = tools.best_arm(list_arm_mu_hat)
if leader_t.nb_times_drawn % self.draw_leader_every == 0:
arm_t = leader_t
else:
if not leader_t.neighbors:
env.set_neighbors(leader_t)
theta_neighbors = []
for cur_neighbor_arm in leader_t.neighbors:
theta_neighbors.append(
(cur_neighbor_arm, self.draw_theta(cur_neighbor_arm)))
arm_t = tools.best_arm(theta_neighbors)
reward_t = arm_t.draw(t)
return arm_t, reward_t, leader_t
def playArm(self, env, t):
#print("list_dle = ", self.list_dle)
assert isinstance(env, UnimodalEnvironment
), "not unimodal environment for unimodal policy"
if t < env.nb_arms:
arm_t = env.get_arm_idx(t)
leader_t = None
else:
list_arm_mu_hat = [(arm, arm.mu_hat) for arm in env.list_of_arms]
leader_t = tools.best_arm(list_arm_mu_hat)
if leader_t.dle_alpha in self.list_dle:
#print("draw leader", leader_t.dle_alpha)
arm_t = leader_t
else:
if not leader_t.neighbors:
env.set_neighbors(leader_t)
theta_neighbors = []
for cur_neighbor_arm in leader_t.neighbors:
theta_neighbors.append(
(cur_neighbor_arm, self.draw_theta(cur_neighbor_arm)))
arm_t = tools.best_arm(theta_neighbors)
leader_t.dle_alpha += 1
if leader_t.dle_alpha == self.y_gamma:
leader_t.dle_alpha = 0
reward_t = arm_t.draw(t)
return arm_t, reward_t, leader_t
def playArm(self, env, #mu_hat_history, t): assert isinstance(env, UnimodalEnvironment), "not unimodal environments" nb_arms = env.nb_arms if t < nb_arms: arm_t = env.get_arm_idx(t) leader_t = None else: list_arm_mu_hat = [(arm, arm.mu_hat) for arm in env.list_of_arms] leader_t = tools.best_arm(list_arm_mu_hat) leader_t.nb_times_leader += 1 if self.draw_leader_every!='no_leader' and leader_t.nb_times_leader%self.draw_leader_every == 0: arm_t = leader_t else: if not leader_t.neighbors: env.set_neighbors(leader_t) UCB_neighbors_idx = [] for cur_neighbor_arm in leader_t.neighbors: UCB_neighbors_idx.append((cur_neighbor_arm, self.KLUCB_idx(cur_neighbor_arm, t))) arm_t = tools.best_arm(UCB_neighbors_idx) reward_t = arm_t.draw(t) return arm_t, reward_t, leader_t
def playArm(self, env, t):
nb_arms = env.nb_arms
if t < nb_arms:
arm_t = env.get_arm_idx(t)
else:
list_arm_UCB = [(arm, self.UCB_idx(arm, t))
for arm in env.list_of_arms]
arm_t = tools.best_arm(list_arm_UCB)
reward_t = arm_t.draw(t)
return arm_t, reward_t
def playArm(self, env, t): nb_arms = env.nb_arms if t < nb_arms: arm_t = env.get_arm_idx(t) else: list_arm_theta = [(arm, self.draw_theta(arm)) for arm in env.list_of_arms] arm_t = tools.best_arm(list_arm_theta) reward_t = arm_t.draw(t) return arm_t, reward_t, leader_t
def playArm(
self,
env,
#mu_hat_history,
t):
assert isinstance(env,
UnimodalEnvironment), "not unimodal environments"
nb_arms = env.nb_arms
if t < nb_arms:
arm_t = env.get_arm_idx(t)
leader_t = None
else:
list_arm_mu_hat = [(arm, arm.mu_hat) for arm in env.list_of_arms]
leader_t = tools.best_arm(list_arm_mu_hat)
print(f"Time {t}, leader is {leader_t.idx}, {leader_t.idx_pair}")
if leader_t.nb_times_drawn % self.draw_leader_every == 0:
arm_t = leader_t
else:
if not leader_t.neighbors:
env.set_neighbors(leader_t)
UCB_neighbors_idx = []
for cur_neighbor_arm in leader_t.neighbors:
print("klucb", self.UCB1_idx(cur_neighbor_arm, t), "arm ",
cur_neighbor_arm.idx)
UCB_neighbors_idx.append(
(cur_neighbor_arm, self.KLUCB_idx(cur_neighbor_arm,
t)))
arm_t = tools.best_arm(UCB_neighbors_idx)
reward_t = arm_t.draw(t)
return arm_t, reward_t, leader_t