Forward–backward algorithm: Difference between revisions

Revision as of 21:42, 16 June 2008

Revision as of 19:05, 2 July 2008

Line 55:

</math>

Note that <math>\mathbf{O_3}</math> differs from the others because of the "no umbrella" observation. Also note that <math>\mathbf{T}</math> and <math>\mathbf{T^T}</math> are equal because the are symmetrical. beforeBefore we can start to compute forward messages we need to describe two special values, the first forward probability and the k+2 backward probability. The first forward probability at t=0 is defined by the prior of the random variable. The k+2 backward probability is defined by the "true" matrix. Therefore ifit follows:

<math>