 Where does Bayesian analysis come in linear regression? Let's say that we have this linear regression with the parameters theta 0, theta 1, and theta 2. If theta 1 and theta 2 are super high, then this is overfitting. Let's say that the mean squared error is represented by this blue dot where corresponding theta 1 value and theta 2 value are super high. We want this value of theta 1 and 2 to be closer to 0. We can now think of a prior distribution for these parameters centered around 0. For example, if we consider a normal distribution prior with the mean 0, we can then use math to get ridge regression. On the other hand, if we consider a Laplace prior centered at 0, we are going to get the formulation for lasso. And so we can derive regularized linear regression equations with a Bayesian lens.