This chapter discusses approximate inference methods for probabilistic models where exact inference is intractable. It introduces variational inference as a deterministic approximation approach. Variational inference works by restricting the distribution of latent variables to a simpler family that makes computation and optimization easier. The chapter provides examples of using variational inference for Gaussian mixtures and univariate Gaussian models. It explains how to derive a variational lower bound and optimize it using an iterative procedure similar to EM.