Base Learner Weights by Inverse Error

Problem

Weight base learners inversely by their errors: w_i = (1/err_i) / sum(1/err_j), clipping each error to at least 1e-12.

Empty input returns [].

Hint

Lower-error learners earn larger ensemble weight.

case 1

in: [[0.1,0.2,0.2]]

expected: [0.5,0.25,0.25]

case 2

in: [[1,1,2]]

expected: [0.4,0.4,0.2]

case 3

in: [[0,1]]

expected: [0.999999999999,9.99999999999e-13]

case 4

in: [[]]

expected: []

Starter codePython

def inverse_error_weights(errors):
    # Your code here
    pass

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