LARS Trust Ratio

Problem

Compute the LARS trust ratio for a weight vector and an update.

Return ||w|| / (||update|| + eps). This scales the update so its norm matches the weight norm.

Hint

eps keeps the ratio finite when the update is the zero vector.

case 1

in: [[3,4],[1,0],1e-8]

expected: 4.999999950000001

case 2

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

expected: 0.5

case 3

in: [[0,0],[1,1],1e-8]

expected: 0

Starter codePython

def lars_trust_ratio(w, update, eps):
    # Your code here
    pass

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