def select(name):
if name.startswith("linear{"):
args = parse_args(name)
n = args["n"]
variance = args["var"]
real_means = linear_means(n)
bandit = linear_bandit(n, variance)
return (real_means, bandit)
elif name.startswith( "polynomial{"):
args = parse_args(name)
n = args["n"]
variance = args["var"]
real_means = polynomial_means(n)
bandit = polynomial_bandit(n, variance)
return (real_means, bandit)
elif name.startswith("linear-linvar{"):
args = parse_args(name)
n = args["n"]
max_variance = args["max_var"]
min_variance = args["min_var"]
real_means = linear_means(n)
bandit = gj_linvar.linear_bandit(n, max_variance, min_variance)
return (real_means, bandit)
elif name.startswith( "polynomial-linvar{"):
args = parse_args(name)
n = args["n"]
max_variance = args["max_var"]
min_variance = args["min_var"]
real_means = polynomial_means()
bandit = gj_linvar.polynomial_bandit(n, max_variance, min_variance)
return (real_means, bandit)
elif name.startswith( "captions{"):
args = parse_args(name)
n = args["n"]
real_means = captions_means(n)
bandit =captions_bandit(n)
return (real_means, bandit)
elif name.startswith( "poisson_olivier{"):
args = parse_args(name)
n = args["n"]
real_means = poisson_oli_means(n)
bandit = poisson_oli_bandit(n)
return (real_means, bandit)
elif name.startswith( "poisson-exp{"):
real_means = poisson_exp_means()
bandit = poisson_exp_bandit()
return (real_means, bandit)
elif name.startswith( "scaled_gaussian{"):
args = parse_args(name)
n = args["n"]
real_means = scaled_gaussian_means()[:n]
bandit = scaled_gaussian_bandit(n)
return (real_means, bandit)
else:
choices = \
["linear{n,var}", "polynomial{n,var}", \
"captions{n}", "poisson-olivier{n}", "scaled_gaussian{n}"]
raise ValueError("Invalid environment \"" + name + "\", " + \
"choose from " + \
"[" + ", ".join(choices) + "]")
def select_means(arm_n, bandit, R_0):
if bandit == "linear":
means_value = linear_means(arm_n)
elif bandit == "sparse":
means_value = sparse_means(arm_n)
elif bandit == "capition":
means_value = captions_means()
elif bandit == "influenza":
means_value = preventive_means(R_0)
else:
means_value = polynomial_means(arm_n)
return means_value
import matplotlib.pyplot as plt
import numpy as np
from algorithms.uniform import Uniform
from environments.gaussian_jun import linear_bandit, linear_means
n = 200
steps = 150000
m = 10
arms_n = 1000
variance = 0.25
# lin_bandits = [linear_bandit(arms_n, variance) for i in range(steps)]
linearmeans = linear_means(arms_n)
J_t_result = np.zeros((2,steps))
for i in range(n):
#lin_bandits = linear_bandit(arms_n, variance, steps)
lin_bandits = linear_bandit(arms_n, variance)
J_t_n = np.zeros((2,steps))
algo = Uniform(lin_bandits, m)
for t in range(steps):
J_t = algo.step(t)
top_m = [linearmeans[int(i)] for i in J_t]
J_t_n[0][t] = np.min(top_m)
J_t_n[1][t] = np.sum(top_m)
J_t_result = J_t_result+J_t_n
result = J_t_result/n
plt.figure(1)
plt.title('Linear m = 10', fontweight='bold')
def linear_bandit(n, max_variance, min_variance):
means = linear_means(n)
variances_ = variances(n, max_variance, min_variance)
return create_bandit(means, variances_)
