enter_evidence.m

来自「贝叶斯网络的matlab实现。可以创建贝叶斯网络、训练模型」· M 代码 · 共 22 行

function [engine, LL] = enter_evidence(engine, ev)
% ENTER_EVIDENCE Call the offline smoother
% [engine, loglik] = enter_evidence(engine, evidence, ...)
%
% evidence{i,t} = [] if if X(i,t) is hidden, and otherwise contains its observed value (scalar or column vector)
%

T = size(ev, 2);
f = cell(1,T);
b = cell(1,T); % b{t}.clpot{c}
ll = zeros(1,T);
[f{1}, ll(1)] = fwd1(engine.tbn_engine, ev(:,1), 1);
for t=2:T
  [f{t}, ll(t)] = fwd(engine.tbn_engine, f{t-1}, ev(:,t), t);
end
LL = sum(ll);
b{T} = backT(engine.tbn_engine, f{T}, T);
for t=T-1:-1:1
  b{t} = back(engine.tbn_engine, b{t+1}, f{t}, t);
end
engine.b = b;