Created
April 28, 2013 22:17
-
-
Save stucchio/5478642 to your computer and use it in GitHub Desktop.
Empirical gain from incorporating priors into the Bayesian Bandit
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from numpy import * | |
from scipy.stats import beta | |
import random | |
class BetaBandit(object): | |
def __init__(self, num_options=2, prior=None): | |
self.trials = zeros(shape=(num_options,), dtype=int) | |
self.successes = zeros(shape=(num_options,), dtype=int) | |
self.num_options = num_options | |
if prior is None: | |
prior = [ (1.0, 1.0) for i in range(num_options)] | |
self.prior = prior | |
def add_result(self, trial_id, success): | |
self.trials[trial_id] = self.trials[trial_id] + 1 | |
if (success): | |
self.successes[trial_id] = self.successes[trial_id] + 1 | |
def get_recommendation(self): | |
sampled_theta = [] | |
for i in range(self.num_options): | |
#Construct beta distribution for posterior | |
dist = beta(self.prior[i][0]+self.successes[i], | |
self.prior[i][1]+self.trials[i]-self.successes[i]) | |
#Draw sample from beta distribution | |
sampled_theta += [ dist.rvs() ] | |
# Return the index of the sample with the largest value | |
return sampled_theta.index( max(sampled_theta) ) | |
prior_params = [(9.0,20.0), (4.0,20.0)] | |
priors = [beta(*x) for x in prior_params] | |
def gain(theta, choice): | |
if (random.random() < theta[choice]): | |
return 1 | |
else: | |
return 0 | |
def gain_bandit(theta, num_trials): | |
bb = BetaBandit(2) | |
g = 0 | |
for i in range(int(num_trials)): | |
choice = bb.get_recommendation() | |
g += gain(theta, choice) | |
return g | |
def gain_prior(theta, num_trials): | |
bb = BetaBandit(2, prior_params) | |
g = 0 | |
for i in range(int(num_trials)): | |
choice = bb.get_recommendation() | |
g += gain(theta, choice) | |
return g | |
num_trials = 50.0 | |
N = 2000 | |
tg = 0 | |
tgb = 0 | |
for i in range(N): | |
theta = [ p.rvs() for p in priors] | |
tg += gain_bandit(theta, num_trials) / num_trials | |
tgb += gain_prior(theta, num_trials) / num_trials | |
print "Base gain: " + str(float(tg)/N) | |
print "Prior gain: " + str(float(tgb)/N) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment