Thompson Sampling Pick

Problem

Pick an arm by Thompson sampling from Beta posteriors.

After random.seed(seed), draw random.betavariate(successes[i] + 1, failures[i] + 1) for every arm and return the index of the largest sample, with the lowest index winning ties.

Hint

Adding 1 to both counts gives a uniform Beta(1, 1) prior per arm.

Test cases

case 1

in: [[1,10,5],[10,1,5],0]

expected: 1

case 2

in: [[0,0,0],[0,0,0],1]

expected: 1

case 3

in: [[9,1],[1,9],42]

expected: 0

case 4

in: [[2,2,2],[8,7,6],7]

expected: 1

Starter codePython

def thompson_sampling_pick(successes, failures, seed):
    # Your code here
    pass

4 test cases · run your code against them in the editor