Logistic Propensity Score

Problem

Return the logistic propensity score 1/(1+exp(-z)) with z = coefficients[0] + sum(coefficients[j+1]*features[j]). Return 0.0 when coefficients does not have exactly len(features) + 1 entries. Handle large negative z without overflow.

Hint

Compute the linear predictor, then squash it through the sigmoid.

Test cases

case 1

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

expected: 0.8175744761936437

case 2

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

expected: 0.5

case 3

in: [[5],[0,-10]]

expected: 1.9287498479639178e-22

case 4

in: [[1],[0]]

expected: 0

case 5

in: [[],[0]]

expected: 0.5

Starter codePython

import math
def logistic_propensity(features, coefficients):
    # Your code here
    pass

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