MediumCalculusca-284

MAE Subgradient Sign Rule

Problem

The mean absolute error has subgradient sign(pred - target) at nonzero residuals and any value in [-1, 1] at zero; use 0 at exact equality.

Given predictions and targets, return the per-example subgradient list with this convention.

Hint

Compare elementwise and return the sign with a zero convention at ties.

case 1

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

expected: [-1,0]

case 2

in: [[0,5,-5],[0,5,5]]

expected: [0,0,-1]

case 3

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

expected: [1,-1]

Starter codePython

def mae_subgradient(preds, targets):
    # Your code here
    pass

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