CMA-ES Mean Update Lite

Problem

Compute the CMA-ES weighted mean update from a ranked population.

Sort indices by fitness descending, with the lower index winning ties. For ranks 0 .. mu-1 use weight log(mu + 0.5) - log(rank + 1), and normalize the selected weights to sum to 1. Return the weighted average of the selected individuals.

Hint

Log-decreasing weights favor the best individuals but never assign zero to the runners-up.

Test cases

case 1

in: [[[0],[2],[4],[10]],[1,4,2,0.5],2]

expected: [2.3916742801345414]

case 2

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

expected: [0.8041628599327295,1.3916742801345412]

case 3

in: [[[5],[1]],[0.1,0.9],1]

expected: [1]

Starter codePython

def cma_mean_update(population, fitness, mu):
    # Your code here
    pass

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