Label Smoothing One-Hot

Problem

Build a smoothed one-hot label vector.

Return a list of num_classes values: (1 - epsilon) + epsilon / num_classes at class_index and epsilon / num_classes at every other position. Round each value to 4 decimal places.

Test cases

case 1

in: [3,0,0.1]

expected: [0.9333,0.0333,0.0333]

case 2

in: [3,1,0]

expected: [0,1,0]

case 3

in: [4,3,0.2]

expected: [0.05,0.05,0.05,0.85]

case 4

in: [2,0,0.5]

expected: [0.75,0.25]

Starter codePython

def label_smoothing(num_classes, class_index, epsilon):
    # Your code here
    pass

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