13.3. Linear Dynamical Systems 637
model for that particular observation. However, the latent variables{zn}are no
longer treated as independent but now form a Markov chain.
Because the model is represented by a tree-structured directed graph, inference
problems can be solved efficiently using the sum-product algorithm. The forward re-
cursions, analogous to theαmessages of the hidden Markov model, are known as the
Kalman filterequations (Kalman, 1960; Zarchan and Musoff, 2005), and the back-
ward recursions, analogous to theβmessages, are known as theKalman smoother
equations, or theRauch-Tung-Striebel(RTS) equations (Rauchet al., 1965). The
Kalman filter is widely used in many real-time tracking applications.
Because the linear dynamical system is a linear-Gaussian model, the joint distri-
bution over all variables, as well as all marginals and conditionals, will be Gaussian.
It follows that the sequence of individually most probable latent variable values is
Exercise 13.19 the same as the most probable latent sequence. There is thus no need to consider the
analogue of the Viterbi algorithm for the linear dynamical system.
Because the model has linear-Gaussian conditional distributions, we can write
the transition and emission distributions in the general form
p(zn|zn− 1 )=N(zn|Azn− 1 ,Γ) (13.75)
p(xn|zn)=N(xn|Czn,Σ). (13.76)
The initial latent variable also has a Gaussian distribution which we write as
p(z 1 )=N(z 1 |μ 0 ,V 0 ). (13.77)
Note that in order to simplify the notation, we have omitted additive constant terms
from the means of the Gaussians. In fact, it is straightforward to include them if
Exercise 13.24 desired. Traditionally, these distributions are more commonly expressed in an equiv-
alent form in terms of noisy linear equations given by
zn = Azn− 1 +wn (13.78)
xn = Czn+vn (13.79)
z 1 = μ 0 +u (13.80)
where the noise terms have the distributions
w ∼N(w| 0 ,Γ) (13.81)
v ∼N(v| 0 ,Σ) (13.82)
u ∼N(u| 0 ,V 0 ). (13.83)
The parameters of the model, denoted byθ={A,Γ,C,Σ,μ 0 ,V 0 }, can be
determined using maximum likelihood through the EM algorithm. In the E step, we
need to solve the inference problem of determining the local posterior marginals for
the latent variables, which can be solved efficiently using the sum-product algorithm,
as we discuss in the next section.