MMD Domain Loss

Problem

Compute the biased maximum mean discrepancy with an RBF kernel.

K(a, b) = exp(-||a - b||^2 / (2 * bandwidth^2)); return mean K(s, s') + mean K(t, t') - 2 * mean K(s, t) rounded to 4 decimal places, averaging over all ordered pairs including self-pairs.

Test cases

case 1

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

expected: 0

case 2

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

expected: 0.7869

case 3

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

expected: 0.5484

Starter codePython

def mmd_rbf(source, target, bandwidth):
    # Your code here
    pass

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