fohria

fohria

import theano
import theano.tensor as tt
import numpy as np
import pymc3 as pm

def update_qvalsQL(action, reward, qvals, alpha, tau, gamma):

    probs = tt.nnet.softmax(qvals * tau)
    probs = probs[0]  # because softmax returns array inside array

fohria

""" just run this until you get a reasonable average for the sequence
    for good arm we use rewards of 2-10 so average should be 6
    for bad arm we use rewards of 1-8 so average should be 4.5 """

import numpy as np

_PRECISION = 0.05 # how far from ideal average our sequence can be
_SEQ_LENGTH = 50 # length of reward sequence
_BAD_ARM_LOW = 1
_BAD_ARM_HIGH = 8

fohria

# define our comma separated sequences of reward values
# average of first 50 is 6.04, average of 51-100 is 4.5
green_rewards = [6, 8, 4, 3, 9, 5, 3, 6, 6, 5, 3, 4, 8, 4, 9, 2, 7, 5, 8, 3, 4, 5, 5,
       3, 9, 3, 7, 8, 8, 9, 7, 6, 9, 4, 8, 9, 9, 8, 8, 5, 8, 3, 7, 7, 7, 3,
       2, 8, 8, 7, 1, 4, 1, 5, 2, 6, 7, 6, 4, 7, 7, 7, 5, 7, 4, 4, 4, 2, 1, 7, 6, 5, 6, 6, 6, 6, 5, 6, 1, 3, 2, 5, 7, 7, 1, 4, 3, 3, 6, 2, 6, 4, 4, 5, 7, 7, 7, 2, 1, 1]
# average of first 50 is 4.46, average of 51-100 is 6.02
orange_rewards = [4, 5, 4, 4, 4, 1, 2, 6, 5, 7, 7, 2, 5, 6, 4, 5, 6, 5, 5, 1, 4, 1, 3, 6, 2, 7, 3, 7, 5, 1, 6, 6, 1, 6, 6, 1, 3, 4, 7, 5, 2, 6, 6, 4, 7, 6, 7, 7, 2, 4, 7, 4, 9, 9, 9, 2, 3, 9, 8, 7, 5, 6, 4, 5, 6, 7, 7, 4, 9, 6, 6, 3, 4,
       7, 7, 9, 7, 5, 4, 6, 5, 8, 8, 3, 2, 6, 3, 7, 5, 2, 8, 8, 9, 3, 4, 7,
       7, 7, 6, 9]

fohria

import numpy as np
import pandas as pd
from scipy.optimize import minimize
import sys
from multiprocessing import Pool
# local modules and functions imported here
sys.path.append('../experiment/simulation')
from utils import softmax, autocorrelation

class MaxLike(object):

fohria

Keybase proof

I hereby claim:

• I am fohria on github.
• I am foh (https://keybase.io/foh) on keybase.
• I have a public key ASBayWEcSsTN-vp0LvAW4683N5MaA_nWHMqQ8z6v-9ksaAo

To claim this, I am signing this object: