Online Gradient Descent Regret Bound

Problem

Compute the standard online gradient descent regret bound.

Return D * G * sqrt(T) for a convex domain of diameter D, a gradient norm bound G, and T rounds. Return 0.0 when T < 1, D <= 0, or G <= 0.

Hint

The bound grows with the square root of the horizon.

case 1

in: [1,2,100]

expected: 20

case 2

in: [0.5,1,10]

expected: 1.5811388300841898

case 3

in: [1,1,0]

expected: 0

case 4

in: [2,0,50]

expected: 0

Starter codePython

def ogd_regret_bound(D, G, T):
    # Your code here
    pass

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