Temperature Move Selection

Problem

Sample a move from temperature-scaled visit counts.

weights[a] = counts[a] ** (1 / temperature); with rng = random.Random(seed), draw u = rng.random() * sum(weights) and return the first index whose cumulative weight exceeds u, falling back to the last index.

Hint

This is inverse-transform sampling over the visit distribution.

Test cases

case 1

in: [[4,1,1],1,0]

expected: 2

case 2

in: [[4,1,1],1,1]

expected: 0

case 3

in: [[4,1,1],1,2]

expected: 2

case 4

in: [[10,0,0],0.5,42]

expected: 0

Starter codePython

import random
def temperature_move_select(counts, temperature, seed):
    # Your code here
    pass

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