wolfenfeld

def run(self):
    """
    run function - Running a game
    """
    # The total rewards of each episode.
    rewards = dict()

    for episode in range(1, self.episodes + 1):

        # The total reward for the current episode.

wolfenfeld

def update_model(self, transition: Transition):
    """
    Updating the module.
    :param transition: state, action, reward, new_state, done
    """
    if transition.reward >= self.best_reward:
        self.best_reward = transition.reward
        self.best_weights = np.copy(self.weights)
        self.alpha = max(self.alpha / 1.5, 1e-2)
    else:

wolfenfeld

def get_action(self, state):
    """
    Getting the action with the highest value for a state
    :param state: the state.
    :return: the action with the highest value for a state.
    """

    return np.argmax(np.dot(state, self.weights))

wolfenfeld

t = 150
mu=1.2
sigma=0.1
alpha_i = sigma * np.random.randn(t) + mu

amount_of_samples = int(0.7*t)

sample_indices = np.random.choice(t, amount_of_samples, replace=False)

sampled_alpha_i = alpha_i[sample_indices]

wolfenfeld

n = 30
mu= 100
sigma=30

x_i = sigma * np.random.randn(n) + mu
t_i = 10*np.ones(n)
v_i = x_i/t_i

amount_of_samples = int(0.7*n)
sample_indices = np.random.choice(n, amount_of_samples, replace=False)

wolfenfeld

n = 50
mu= 100
sigma=40

t_i = sigma * np.random.randn(n) + mu
x_i = 1000*np.ones(n)
v_i = x_i/t_i

amount_of_samples = int(0.7*n)
sample_indices = np.random.choice(n, amount_of_samples, replace=False)

wolfenfeld

import numpy as np

from scipy.stats import gmean,hmean

wolfenfeld

class BanditSimulation(object):
    
    def __init__(self, k_arms, T, number_of_runs):
        self.T = T
        self.number_of_runs = number_of_runs
        self.k_arms = k_arms
        self.cumulative_regret = {"Thompson Sampling":np.zeros((T,number_of_runs)),
                                  "UCB":np.zeros((T,number_of_runs))}
    
    def run(self):

wolfenfeld

class BernoulliArm(object):

    def __init__(self, mu):
        self.mu = mu
  
    def draw(self):
        return np.random.binomial(n=1, p=self.mu)   

wolfenfeld

class BanditAlgorithm(object):
    
    def __init__(self, k_arms):
        pass
    
    def select_arm(self):
        pass
    
    def update(self, chosen_arm, reward):
        pass