matlab/total_model_log_likelihood.m

   1 function log_likelihood = total_model_log_likelihood(X, model)\r
   2 n = size(X.obs, 2);\r
   3 log_likelihood_vect = zeros(n, 1);\r
   4 for i=1:size(X.obs, 2)\r
   5    log_likelihood_vect(i) = log(model_likelihood(X.obs(i).obs, model));\r
   6 end\r
   7 log_likelihood = sum(log_likelihood_vect);\r
   8 \r
   9 % To compute the model likelihood for a single observation, we add together\r
  10 % the probabilities of each of the underlying Markov Chains multiplied by\r
  11 % their priors.\r
  12 %function likelihood = model_likelihood(observation, model)\r
  13 %likelihood = 0;\r
  14 %likelihood_vect = zeros(model.num_mcs, 1);\r
  15 %for i=1:model.num_mcs\r
  16 %    likelihood_vect(i) = model.priors(i) * chain_likelihood(observation, model.chain(i));\r
  17 %end\r
  18 %likelihood_vect;\r
  19 %likelihood = sum(likelihood_vect);\r
  20 \r
  21 % The likelihood of the observation for a single Markov Chain\r
  22 %function likelihood = chain_likelihood(observation, mc)\r
  23 %likelihood = mc.t(observation(1));\r
  24 %for i=2:size(observation, 2)\r
  25 %    likelihood = likelihood * mc.T(observation(i-1), observation(i));\r
  26 %end