annotate toolboxes/FullBNT-1.0.7/bnt/inference/dynamic/@kalman_inf_engine/kalman_inf_engine.m @ 0:e9a9cd732c1e
tip
first hg version after svn
| author |
wolffd |
| date |
Tue, 10 Feb 2015 15:05:51 +0000 |
| parents |
|
| children |
|
| rev |
line source |
|
wolffd@0
|
1 function engine = kalman_inf_engine(bnet)
|
|
wolffd@0
|
2 % KALMAN_INF_ENGINE Inference engine for Linear-Gaussian state-space models.
|
|
wolffd@0
|
3 % engine = kalman_inf_engine(bnet)
|
|
wolffd@0
|
4 %
|
|
wolffd@0
|
5 % 'onodes' specifies which nodes are observed; these must be leaves.
|
|
wolffd@0
|
6 % The remaining nodes are all hidden. All nodes must have linear-Gaussian CPDs.
|
|
wolffd@0
|
7 % The hidden nodes must be persistent, i.e., they must have children in
|
|
wolffd@0
|
8 % the next time slice. In addition, they may not have any children within the current slice,
|
|
wolffd@0
|
9 % except to the observed leaves. In other words, the topology must be isomorphic to a standard LDS.
|
|
wolffd@0
|
10 %
|
|
wolffd@0
|
11 % There are many derivations of the filtering and smoothing equations for Linear Dynamical
|
|
wolffd@0
|
12 % Systems in the literature. I particularly like the following
|
|
wolffd@0
|
13 % - "From HMMs to LDSs", T. Minka, MIT Tech Report, (no date), available from
|
|
wolffd@0
|
14 % ftp://vismod.www.media.mit.edu/pub/tpminka/papers/minka-lds-tut.ps.gz
|
|
wolffd@0
|
15
|
|
wolffd@0
|
16 [engine.trans_mat, engine.trans_cov, engine.obs_mat, engine.obs_cov, engine.init_state, engine.init_cov] = ...
|
|
wolffd@0
|
17 dbn_to_lds(bnet);
|
|
wolffd@0
|
18
|
|
wolffd@0
|
19 % This is where we will store the results between enter_evidence and marginal_nodes
|
|
wolffd@0
|
20 engine.one_slice_marginal = [];
|
|
wolffd@0
|
21 engine.two_slice_marginal = [];
|
|
wolffd@0
|
22
|
|
wolffd@0
|
23 engine = class(engine, 'kalman_inf_engine', inf_engine(bnet));
|
