Most machine learning algorithms are derived from some statistical models. The statistical model helps analyze the theoretical properties of a machine learning algorithm. Which of the following statements is (are) true? (10 points)

1)The widely used least squares method is a maximum likelihood approach with a Gaussian distribution assumption on residuals (or sometimes called data observation noise).

2)The widely used least squares method is a maximum likelihood approach with a Bernoulli distribution assumption on residuals.

3)A regularized regression method is a maximum a posterior approach with a Gaussian distribution assumption on residuals and a Gaussian prior on w.

4)The ridge regression method is a maximum likelihood approach with a Gaussian distribution assumption on residuals and a Gaussian prior on w.

5) The LASSO method is a maximum a posterior approach with a Laplace prior distribution on w.