ReLU With Gradient

Problem

Apply ReLU and return the forward values plus local gradients.

Return [outputs, grads] where outputs[i] = max(0, x_i) and grads[i] is 1.0 when the input is strictly positive, otherwise 0.0. An empty input returns [[], []].

Hint

The derivative of ReLU is a step function: 1 above zero, 0 elsewhere.

Test cases

case 1

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

expected: [[0,0,2],[0,0,1]]

case 2

in: [[5]]

expected: [[5],[1]]

case 3

in: [[]]

expected: [[],[]]

case 4

in: [[-3,-0.5]]

expected: [[0,0],[0,0]]

Starter codePython

def relu_with_grad(x):
    # Return [outputs, grads]
    # Your code here
    pass

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