Linear dynamical system. This set of functions performs inference and learning of a linear Kalman filter model. Inference is carried out via forward-backward smoothing, and learning is accomplished via the expectation maximization algorithm.