Knowledge MaxxingMean squared error
Mean squared error is the average of the squared differences between predicted and true values, which penalizes large errors more heavily than small ones.
Prerequisite: Mean absolute error
What matters
Squaring the errors before averaging means one very large mistake contributes disproportionately more than several small mistakes, which shapes which errors a model is pushed to avoid during training.
Example
A single $100,000 error contributes far more to mean squared error than ten separate $10,000 errors, even though the total error size is similar.
Common misconception
A lower mean squared error does not always mean a model is better in every practical sense; it specifically reflects a preference for avoiding large errors over many small ones.
Recall question
Why does one large error affect mean squared error more than several small ones?