# The basic pattern is 1 look at the current partition

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: (1, t )) ∗ maxq∈{q∗ (1),q∗ (2)} mt −1,t (qt −1 (2), q ) t t ∗ ∗ mt −1,t (qt −1 (2), qt (2)) (9.28) and for t &gt; t(2) = t pm DQ1:T (1, t , t) = pm (DQ1:T (1, t )) ∗ maxq=qt (2) mt−1,t (qt−1 (2), q ) ∗ ∗ mt−1,t (qt−1 (2), qt (2)) (9.29) the third most probable sequence has probability ∗ p(q1:T (3), x1:T ) = max max pm (DQ1:T (1, t)), max pm (DQ1:T (1, t , t)) ¯ t:t=t t:t≥t (9.30) Given new best block, use dynamic programming again to identify it. Prof. Jeﬀ Bilmes EE596A/Winter 2013/DGMs – Lecture 9 - Feb 6th, 2013 page 9-25 (of 180) k-best without the penalty Island Summary Scratch General case the same pattern as above, iterative procedure to construct the next best. Prof. Jeﬀ Bilmes EE596A/Winter 2013/DGMs – Lecture 9 - Feb 6th, 2013 page 9-26 (of 180) k-best without the penalty Island Summary Scratch General case the same pattern as above, iterative procedure to construct the next best. The basic pattern is: Prof. Jeﬀ Bilmes EE596A/Winter 2013/DGMs – Lectur...
View Full Document

## This document was uploaded on 04/05/2014.

Ask a homework question - tutors are online