Mercurial > hg > camir-aes2014
diff toolboxes/FullBNT-1.0.7/bnt/inference/static/@gibbs_sampling_inf_engine/gibbs_sampling_inf_engine.m @ 0:e9a9cd732c1e tip
first hg version after svn
author | wolffd |
---|---|
date | Tue, 10 Feb 2015 15:05:51 +0000 |
parents | |
children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/toolboxes/FullBNT-1.0.7/bnt/inference/static/@gibbs_sampling_inf_engine/gibbs_sampling_inf_engine.m Tue Feb 10 15:05:51 2015 +0000 @@ -0,0 +1,104 @@ +function engine = gibbs_sampling_inf_engine(bnet, varargin) +% GIBBS_SAMPLING_INF_ENGINE +% +% engine = gibbs_sampling_inf_engine(bnet, ...) +% +% Optional parameters [default in brackets] +% 'burnin' - How long before you start using the samples [100]. +% 'gap' - how often you use the samples in the estimate [1]. +% 'T' - number of samples [1000] +% i.e, number of node flips (so, for +% example if there are 10 nodes in the bnet, and T is 1000, each +% node will get flipped 100 times (assuming a deterministic schedule)) +% The total running time is proportional to burnin + T*gap. +% +% 'order' - if the sampling schedule is deterministic, use this +% parameter to specify the order in which nodes are sampled. +% Order is allowed to include multiple copies of nodes, which is +% useful if you want to, say, focus sampling on particular nodes. +% Default is to use a deterministic schedule that goes through the +% nodes in order. +% +% 'sampling_dist' - when using a stochastic sampling method, at +% each step the node to sample is chosen according to this +% distribution (may be unnormalized) +% +% The sampling_dist and order parameters shouldn't both be used, +% and this will cause an assert. +% +% +% Written by "Bhaskara Marthi" <bhaskara@cs.berkeley.edu> Feb 02. + + +engine.burnin = 100; +engine.gap = 1; +engine.T = 1000; +use_default_order = 1; +engine.deterministic = 1; +engine.order = {}; +engine.sampling_dist = {}; + +if nargin >= 2 + args = varargin; + nargs = length(args); + for i = 1:2:nargs + switch args{i} + case 'burnin' + engine.burnin = args{i+1}; + case 'gap' + engine.gap = args{i+1}; + case 'T' + engine.T = args{i+1}; + case 'order' + assert (use_default_order); + use_default_order = 0; + engine.order = args{i+1}; + case 'sampling_dist' + assert (use_default_order); + use_default_order = 0; + engine.deterministic = 0; + engine.sampling_dist = args{i+1}; + otherwise + error(['unrecognized parameter to gibbs_sampling_inf_engine']); + end + end +end + +engine.slice_size = size(bnet.dag, 2); +if (use_default_order) + engine.order = 1:engine.slice_size; +end +engine.hnodes = []; +engine.onodes = []; +engine.evidence = []; +engine.state = []; +engine.marginal_counts = {}; + +% Precompute the strides for each CPT +engine.strides = compute_strides(bnet); + +% Precompute graphical information +engine.families = compute_families(bnet); +engine.children = compute_children(bnet); + +% For convenience, store the CPTs as tables rather than objects +engine.CPT = get_cpts(bnet); + +engine = class(engine, 'gibbs_sampling_inf_engine', inf_engine(bnet)); + + + + + + + + + + + + + + + + +