Melvin Leok
12 minutes
Least squares chooses parameters that minimise the sum of squared residuals between observed and predicted values, giving a closed-form solution for linear models. It implicitly assumes errors are independent, of constant variance, and concentrated in the response variable.