A measure of the average squared difference between predicted values and observed values, widely used to quantify estimation and prediction error in regression and statistical learning.
Semantic Classification
Content
- Mean squared error averages the squared residuals between predictions and targets, penalising larger errors disproportionately. It decomposes into bias and variance components, linking it directly to statistical learning theory.
- As a differentiable loss function it is convenient for optimisation by gradient descent and is a default choice for regression tasks in supervised learning.