comparison toolboxes/FullBNT-1.0.7/bnt/inference/static/@gibbs_sampling_inf_engine/private/compute_strides.m @ 0:e9a9cd732c1e tip

first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 function strides = compute_strides(bnet)
2 % COMPUTE_STRIDES For each CPT and each variable in that CPT,
3 % returns the stride of that variable. So in future, we can
4 % quickly extract a slice of the CPT.
5 %
6 % The return value is a 2d array, where strides(i,j) contains the
7 % stride of the jth variable in the ith CPT. Cell arrays would
8 % have saved space but they are slower.
9 %
10
11 num_cpts = size(bnet.CPD, 2);
12 max_cpt_dim = 1 + max(sum(bnet.dag));
13 strides = zeros(num_cpts, max_cpt_dim);
14
15 for i = 1:num_cpts
16 c = CPT(bnet, i);
17 siz = size(CPT(bnet, i));
18
19 % Deal with the special case of a 1-d array separately
20 if siz(2) == 1
21 dim = 1;
22 else
23 dim = size(siz, 2);
24 end
25
26 strides(i, 1:dim ) = [1 cumprod(siz(1:dim-1))];
27 end