Evolution Strategies Gradient Estimate

Problem

Estimate the evolution strategies gradient from sampled perturbations: (1 / (n * sigma)) * sum of fitness[i] * epsilon[i], returned as a vector. An empty batch returns [].

Hint

The score-function trick turns black-box fitness into a gradient estimate.

Test cases

case 1

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

expected: [0,0]

case 2

in: [[2,4],[[1,-1],[0,1]],0.5]

expected: [2,2]

case 3

in: [[1,2,3],[[1,0],[0,1],[-1,-1]],2]

expected: [-0.3333333333333333,-0.16666666666666666]

Starter codePython

def es_gradient_estimate(fitness, epsilons, sigma):
    # Your code here
    pass

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