Model Soup Mean And Spread

Problem

Average a list of model parameter vectors and measure their spread.

All models are flat lists of the same length. Return [mean, max_deviation] where mean is the coordinatewise average and max_deviation is the largest absolute difference between any model coordinate and its mean. An empty list returns [[], 0.0].

Hint

The spread tells you how far the soup ingredients are from the final average.

Test cases

case 1

in: [[[1,2],[3,4],[5,6]]]

expected: [[3,4],2]

case 2

in: [[[0,0],[2,4]]]

expected: [[1,2],2]

case 3

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

expected: [[1],0]

case 4

in: [[]]

expected: [[],0]

Starter codePython

def model_soup_mean_and_spread(models):
    # Returns [mean, max_deviation]
    # Your code here
    pass

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