HMM alg purpose: Forward algorithm/Viterbi algorithm/BaumWelch algorithm

Author 
Message 
g3yangs
Student
Reputation: 0

HMM alg purpose: Forward algorithm/Viterbi algorithm/BaumWelch algorithm
Hey all,
Could someone explain what are the purposes for those three different algs(Forward algorithm/Viterbi algorithm/BaumWelch algorithm) for HMM?
thanks


20160417 15:45 

g3kooks
Student
Reputation: 1

RE: HMM alg purpose: Forward algorithm/Viterbi algorithm/BaumWelch algorithm
(20160417 15:45)g3yangs Wrote: Hey all,
Could someone explain what are the purposes for those three different algs(Forward algorithm/Viterbi algorithm/BaumWelch algorithm) for HMM?
thanks
forward procedure allows you to determine the likelihood of a sequence of observations up until and including time t.
backward procedure is the same, except it's the sequence of observations beyond a time t up until the end of the sequence at time T.
if you put forward and backward together, you can determine the likelihood of a sequence at any given time t.
the viterbi algorithm allows you to actually find the sequence of states that maximize likelihood. it yields a path from some node in the trellis at time 0 to some node in the trellis at time T.
baumwelch is the expectationmaximization training algorithm tailored towards HMMs. you initialize your forward and backward values, and then you compute the expectation of 1) being in a specific state at time t given the entire observation sequence, and 2) transitioning from one state to another at time t given the entire observation sequence and the parameters you compute in 1). after you've computed those, you use MLE formulas to maximize their values given the current parameters.


20160417 16:12 

Thread Rating:
 0 Votes  0 Average
 1
 2
 3
 4
 5
User(s) browsing this thread: 1 Guest(s)