At this point we have seen that models can be too complicated, but thus far complexity has roughly meant the number of parameters (regression coefficients). Striking a balance between bias and variance means finding the right number of predictors to include, but there are ways to build highly flexibly models which do not overfit.s
Material:
Regularization for regression models, Priors, Laplace rule of succession, Bayes rule, posterior distributions, how priors influence the posterior for simple models, connection between regularization and priors.