AkihikoITOH/simab

from simab.arms.normal import NormalArm
from simab.algorithms.epsilon_first import EpsilonFirst
from simab.algorithms.epsilon_greedy import EpsilonGreedy
from simab.algorithms.oracle import Oracle
from simab.algorithms.random_choice import Random
from simab.algorithms.single_arm import Single
from simab.algorithms.softmax import Softmax
from simab.algorithms.ucb1 import UCB1

ROUNDS = 1000

# Generate five ND-arms.
arms = [NormalArm(0.1*float(i), 0.1) for i in range(3, 8)]

# Generate an agent.
algorithm = Softmax(arms, 0.1)
# algorithm = EpsilonFirst(arms, 0.1, rounds=ROUNDS)
# algorithm = EpsilonGreedy(arms, 0.1)
# algorithm = UCB1(arms)
# algorithm = Single(arms, 2)
# algorithm = Random(arms)

for _ in range(ROUNDS):
    algorithm.play()

summary = algorithm.summary()
print summary['algorithm']
print summary['tau']
print summary['plays']
print summary['total_reward']
print summary['true_means']
print summary['true_sds']
print summary['empirical_means']
print summary['empirical_sds']
# print summary['history']