# Eg here is a t j pqt j x1t backwards recursion

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: Summary Scratch HMM, other recursions It is possible to derive a temporal recursion for quantities other than α and β . E.g., here is a γt (j ) = p(Qt = j |x1:T ) backwards recursion. γ ( qt ) = qt+1 = qt+1 = qt+1 Prof. Jeﬀ Bilmes p(qt , qt+1 |x1:T ) = qt+1 p(qt |qt+1 , x1:T )p(qt+1 |x1:T ) p(qt |qt+1 , x1:T )γ (qt+1 ) = qt+1 p(qt |qt+1 , x1:t )γ (qt+1 ) p(qt , qt+1 , x1:t ) γ (qt+1 ) q p(qt , qt+1 , x1:t ) t EE596A/Winter 2013/DGMs – Lecture 5 - Jan 25th, 2013 page 5-21 (of 232) HMMs Trellis Other HMM queries MPE Sampling What HMMs can do Summary Scratch HMM, other recursions It is possible to derive a temporal recursion for quantities other than α and β . E.g., here is a γt (j ) = p(Qt = j |x1:T ) backwards recursion. γ ( qt ) = qt+1 = qt+1 p(qt , qt+1 |x1:T ) = qt+1 p(qt |qt+1 , x1:T )p(qt+1 |x1:T ) p(qt |qt+1 , x1:T )γ (qt+1 ) = qt+1 p(qt |qt+1 , x1:t )γ (qt+1 ) qt+1 p(qt , qt+1 , x1:t ) γ (qt+1 ) q p(qt , qt+1 , x1:t ) qt+1 p(qt+1 |qt )p(qt , x1:t ) γ (qt+1 ) q p(qt+1 |qt )p(qt , x1:t ) = = Prof. Jeﬀ Bilmes t t EE596A/Winter 2013/DGMs – Lecture 5 - Jan 25th, 2013 page 5-21 (of 232) HMMs Trellis Other HMM queries MPE Sampling What HMMs can do Summary Scratch HMM, other recursions It is possible to derive a temporal recursion for quantities other than α and β . E.g., here is a γt (j ) = p(Qt = j |x1:T ) backwards recursion. γ ( qt ) = qt+1 = qt+1 p(qt , qt+1 |x1:T ) = qt+1 p(qt |qt+1 , x1:T )p(qt+1 |x1:T ) p(qt |qt+1 , x1:T )γ (qt+1 ) = qt+1 p(qt |qt+1 , x1:t )γ (qt+1 ) qt+1 p(qt , qt+1 , x1:t ) γ (qt+1 ) q p(qt , qt+1 , x1:t ) qt+1 p(qt+1 |qt )p(qt , x1:t ) γ (qt+1 ) q p(qt+1 |qt )p(qt , x1:t ) qt+1 p(qt+1 |qt )αt (qt ) γ (qt+1 ) q p(qt+1 |qt )αt (qt ) = = = Prof. Jeﬀ Bilmes t t t EE596A/Winter 2013/DGMs – Lecture 5 - Jan 25th, 2013 page 5-21 (of 232) HMMs Trellis Other HMM queries MPE Sampling What HMMs can do Summary Scratch HMM, other recursions It is possible to derive a temporal recursion for quantities other than α and β . E.g., here is a γt (j ) = p(Qt = j |x1:T ) backwards recursion. γ ( qt ) = qt+1 = qt+1 p(qt , qt+1 |x1:T ) = qt+1 p(qt |qt+1 , x1:T )p(qt+1 |x1:T ) p(qt |qt+1 , x1:T )γ (qt+1 ) = qt+1 p(qt |qt+1 , x1:t )γ (qt+1 ) qt+1 p(qt , qt+1 , x1:t ) γ (qt+1 ) q p(qt , qt+1 , x1:t ) qt+1 p(qt+1 |qt )p(qt , x1:t ) γ (qt+1 ) q p(qt+1 |qt )p(qt , x1:t ) qt+1 p(qt+1 |qt )αt (qt ) γ (qt+1 ) q p(qt+1 |qt )αt (qt ) = = = t t t Therefore, there is a backwards pass recursion using just the α’s without directly touching the observations again (better memory). Prof. Jeﬀ Bilmes EE596A/Winter 2013/DGMs – Lecture 5 - Jan 25th, 2013 page 5-21 (of 232) HMMs Trellis Other HMM queries MPE Sampling What HMMs can do Summary Scratch HMM Trellis/Lattice/Grid We’ll see that it is very useful to view the HMM state-space as a trellis. Prof. Jeﬀ Bilmes EE596A/Winter 2013/DGMs – Lecture 5 - Jan 25th, 2013 page 5-22 (of 232) HMMs Trellis Other HMM queries MPE Sampling What HMMs can do Summary Scratch HMM Trellis/Lattice/Grid We’ll see that it is very useful to view the HMM state-space as a trellis. It is sometimes called an HMM grid, or even a “...
View Full Document

Ask a homework question - tutors are online