Adam Moment Estimate Update

Problem

Adam maintains m_new = beta1 * m + (1 - beta1) * g and v_new = beta2 * v + (1 - beta2) * g^2.

Given the current m, v, the gradient g, beta1, and beta2, return [m_new, v_new] for a scalar parameter.

Hint

The first moment averages gradients, the second averages squared gradients.

case 1

in: [0,0,1,0.9,0.999]

expected: [0.09999999999999998,0.0010000000000000009]

case 2

in: [1,2,-0.5,0.5,0.5]

expected: [0.25,1.125]

case 3

in: [0.2,0.1,3,0.99,0.999]

expected: [0.22800000000000004,0.10890000000000001]

Starter codePython

def adam_moment_estimate_update(m, v, g, beta1, beta2):
    # Your code here
    pass

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