Self-Play Temperature Schedule

Problem

Compute the move-selection temperature during self-play.

Before switch_move the temperature is 1.0. Afterwards it decays linearly, tau = 1 - (move_index - switch_move) * (1 - tau_min) / decay_steps, floored at tau_min. A non-positive decay_steps jumps straight to tau_min.

Hint

Early moves stay exploratory; late moves become nearly greedy.

Test cases

case 1

in: [0,10,20,0.1]

expected: 1

case 2

in: [10,10,20,0.1]

expected: 1

case 3

in: [20,10,20,0.1]

expected: 0.55

case 4

in: [50,10,20,0.1]

expected: 0.1

case 5

in: [15,10,0,0.5]

expected: 0.5

Starter codePython

def self_play_temperature_schedule(move_index, switch_move, decay_steps, tau_min):
    # Your code here
    pass

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