Gradient Variance Estimate

Problem

Estimate the mean and variance of a batch of scalar gradient samples.

Return [mean, population_variance] using division by n. An empty batch returns [0.0, 0.0].

Hint

This estimates how noisy the policy-gradient estimator is.

case 1

in: [[1,3]]

expected: [2,1]

case 2

in: [[2,2,2]]

expected: [2,0]

case 3

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

expected: [0,0.6666666666666666]

case 4

in: [[]]

expected: [0,0]

Starter codePython

def gradient_variance(grads):
    # Your code here
    pass

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