Sensitivity at Specificity

Problem

Find the threshold with the highest sensitivity subject to specificity >= target_specificity. Sweep unique probabilities from high to low.

Return [threshold, sensitivity]; ties keep the higher threshold, and [0.0, 0.0] when nothing qualifies.

Hint

Trading specificity for sensitivity traces the ROC curve from the conservative end.

case 1

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],1]

expected: [0.8,0.5]

case 2

in: [[0,0,1,1],[0.1,0.4,0.35,0.8],0.5]

expected: [0.35,1]

case 3

in: [[1,1],[0.2,0.8],1]

expected: [0,0]

case 4

in: [[],[],1]

expected: [0,0]

Starter codePython

def sensitivity_at_specificity(y_true, probs, target_specificity):
    # Returns [threshold, sensitivity]
    # Your code here
    pass

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