Huber Loss

Problem

Compute the mean Huber loss with parameter delta. For residual r = y - pred use 0.5*r^2 when |r| <= delta, otherwise delta*(|r| - 0.5*delta).

Empty input returns 0.0.

Hint

Quadratic for small residuals, linear beyond delta so outliers do not dominate.

case 1

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

expected: 0

case 2

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

expected: 2.5

case 3

in: [[0,0],[0.5,3],2]

expected: 2.0625

case 4

in: [[],[],1]

expected: 0

Starter codePython

def huber_loss(y_true, y_pred, delta):
    # Your code here
    pass

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