This image shows the formula for the Mean Absolute Error, also known as MAE. First the sum is taken, for values from 1 to n, of the absolute, signless differences between each actual likelihood and the corresponding predicted likelihood. Then the sum is divided by n.