EasyCalculusca-283

Huber Loss Gradient

Problem

For residuals r = pred - target, the Huber loss uses gradient r when |r| <= delta and delta * sign(r) otherwise.

Given predictions, targets, and delta, return the gradient list.

Hint

Huber is quadratic near zero and linear outside the delta band.

case 1

in: [[1,5],[2,2]]

expected: [-1,1]

case 2

in: [[1.5,1.5],[0,3],1]

expected: [1,-1]

case 3

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

expected: [2,-2]

Starter codePython

def huber_gradient(preds, targets, delta=1.0):
    # Your code here
    pass

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