Centralized Gradient Norm

Problem

Compute the Euclidean norm of a gradient vector after gradient centralization.

Subtract the mean of the components from each component, then return the norm of the centered vector. An empty input returns 0.0, and a constant vector centers to the zero vector.

Hint

Centralization removes the projection onto the all-ones direction.

Test cases

case 1

in: [[1,2,3]]

expected: 1.4142135623730951

case 2

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

expected: 2.23606797749979

case 3

in: [[4,4,4]]

expected: 0

case 4

in: [[]]

expected: 0

case 5

in: [[5]]

expected: 0

Starter codePython

def centralized_grad_norm(g):
    # Your code here
    pass

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