Inverse-Frequency Class Weights

Problem

Compute inverse-frequency class weights. For each label c, weight = n / (k * count_c), where n is the number of samples and k is the number of distinct labels.

Return a dict mapping str(label) to its weight. Empty input returns {}.

Hint

Rarer classes get larger weights; keys are stringified labels.

Test cases

case 1

in: [["a","a","a","b","b","c"]]

expected: {"a":0.6666666666666666,"b":1,"c":2}

case 2

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

expected: {"0":1.5,"1":0.75}

case 3

in: [["x"]]

expected: {"x":1}

case 4

in: [[]]

expected: {}

Starter codePython

def inverse_frequency_weights(labels):
    # Your code here
    pass

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