L1 Regularized Subgradient

Problem

Adding an L1 penalty lambda * ||w||_1 adds lambda * sign(w) to the gradient, using the subgradient 0 at w = 0.

Given the data gradient and the weights, return the regularized subgradient.

Hint

The subgradient of |w| is the sign of w, with zero chosen at the kink.

case 1

in: [[1,2],[3,4],0.5]

expected: [1.5,2.5]

case 2

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

expected: [0,-1,0]

case 3

in: [[-1],[5],0.25]

expected: [-0.75]

Starter codePython

def l1_regularized_subgradient(grad, weights, lam):
    # Your code here
    pass

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