Is it k-fold (normally 5 or 10) or Leave-one-out cross-validation (LOOCV). K-fold is the commonly chosen option.
Method for linear and other regression forms to limit the number of parameters we use to model the system. The more parameters, the worse the weighting of the parameters. Drives models to be more simple
An improvement on Ridge Regression to minimise the number of predictors used to build a model - making it a sparse rather than dense model