In [1]:

    

%matplotlib inline
import matplotlib

import tanuki
from tanuki.bandit import GaussianBandit
from tanuki.agent import Agent, GradientAgent
from tanuki.policy import EpsilonGreedyPolicy
from tanuki.environment import Environment


    

In [2]:

    

n_arms = 10
bandit = GaussianBandit(n_arms)
n_trials = 1000
n_experiments = 500


    

In [6]:

    

#agents = [Agent(bandit, EpsilonGreedyPolicy(0.01))]

policy = EpsilonGreedyPolicy(0.01)

agents = [GradientAgent(bandit, policy, alpha=0.1),
          GradientAgent(bandit, policy, alpha=0.8)]

env = Environment(bandit, agents, 'Epsilon-Greedy')
scores, optimal = env.run(n_trials, n_experiments)


    

In [4]:

    

matplotlib.rcParams['figure.figsize'] = [10.0, 10.0]

env.plot_results(scores, optimal)


    

In [7]:

    

env.plot_beliefs()


    

In [ ]: