Entropy Collapse Check

Problem

Detect entropy collapse: return True when the last patience entropies are all below ratio * entropies[0], and False when there are fewer entries, patience is non-positive, or any of the last entries is at or above the threshold.

Hint

A policy whose entropy keeps sinking is becoming deterministic and losing exploration.

Test cases

case 1

in: [[1,0.5,0.2,0.1],0.3,2]

expected: true

case 2

in: [[1,0.5,0.2],0.3,2]

expected: false

case 3

in: [[1],0.5,1]

expected: false

case 4

in: [[],0.5,1]

expected: false

Starter codePython

def entropy_collapse(entropies, ratio, patience):
    # Your code here
    pass

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