MAML Inner Step

Problem

Run one MAML inner-loop gradient step on a linear model.

With prediction sum(w * x) and loss 0.5 * (pred - y)^2, the gradient is (pred - y) * x per weight. Return [adapted_weights, round(loss, 4)] where adapted_weights = w - lr * grad rounded to 4 decimal places.

Test cases

case 1

in: [[1,0],[2,1],0,0.1]

expected: [[0.6,-0.2],2]

case 2

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

expected: [[0.5],0]

case 3

in: [[1,1],[1,1],3,0.2]

expected: [[1.2,1.2],0.5]

Starter codePython

def maml_inner_step(weights, x, y, lr):
    # Your code here
    pass

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