masced-bandits-EGAlberts 0.0.1

bandit algs

MASCed_bandits

This is a library of multi-armed bandit policies. As of the most recent version the following policies are included: UCB, UCB-Improved, EXP3, EXP3S, EXP4, EwS, ETC, Discounted UCB, Sliding Window UCB, e-greedy.

Example

from masced_bandits.bandit_options import initialize_arguments
from masced_bandits.bandits import init_bandit
import numpy as np

initialize_arguments(["Arm1","Arm2"], 0)

ucb_instance = init_bandit(name='UCB')
for i in range(100):
    arms_chosen = []
    reward = np.random.random()
    arms_chosen.append(ucb_instance.get_next_arm(reward))