Warm-up loss · training

Mean squared error

The one number a model is trying to push down while it learns.

When a model guesses numbers, we need one score for how wrong it is. Mean squared error takes each miss, squares it so big misses count extra, then averages.

Write mse(predicted, actual): two arrays of numbers with the same length. Return the average of (predicted[i] - actual[i]) squared. If the arrays are empty, return 0.

Example: mse([1, 2], [1, 4]) returns 2, because the misses are 0 and 2, the squares are 0 and 4, and their average is 2.

Tests you can see

mse([1,2], [1,4])2
mse([3,5,7], [3,5,7])0
mse([2,0,4], [1,1,1])3.6666666666666665

More tests are hidden. The server runs them all before a time counts.

The idea behind this, in a two-minute live page

Run your code to see the tests.

javascriptmse()

function mse(predicted, actual) {
  // average of the squared misses
  return 0;
}

function mse(predicted, actual) {
  // average of the squared misses
  return 0;
}

Tab indents · Cmd/Ctrl+Enter runs