matlab/generate_sequences.m

   1 % Generate sample sequences from the model given.\r
   2 function sequences = generate_sequences(model, num_sequences, sequence_length)\r
   3 \r
   4 for i=1:num_sequences\r
   5     sequences(i).obs = generate_sequence_from_model(model, sequence_length);\r
   6 end\r
   7 \r
   8 function sequence = generate_sequence_from_model(model, sequence_length)\r
   9 \r
  10 % Figure out which MC we want to use\r
  11 which_mc = get_random_index(model.priors');\r
  12 sequence = generate_sequence_from_mc(model.chain(which_mc), sequence_length);\r
  13 \r
  14 function sequence = generate_sequence_from_mc(mc, sequence_length)\r
  15 \r
  16 sequence = zeros(1, sequence_length);\r
  17 \r
  18 % First generate the initial state\r
  19 sequence(1) = get_random_index(mc.t');\r
  20 % Now generate the next states based on the transition matrices\r
  21 for i=2:sequence_length\r
  22     sequence(i) = get_random_index(mc.T(sequence(i-1), :));\r
  23 end\r
  24 \r
  25 % Returns a random index, assuming that the passed in vector is a 1 x n\r
  26 % vector whose contents sum to 1.\r
  27 function index = get_random_index(dist)\r
  28 assert(abs(sum(dist) - 1) < 0.001);\r
  29 sequence_rand = rand(1);\r
  30 current_min = 0;\r
  31 index = -1;\r
  32 for i=1:size(dist, 2)\r
  33     current_max = current_min + dist(i);\r
  34     if (sequence_rand >= current_min & sequence_rand < current_max)\r
  35         index = i;\r
  36         break;\r
  37     end\r
  38     current_min = current_max;\r
  39 end\r
  40 assert(index > 0 & index <= size(dist, 2))\r