HMM: Three Problems• What is the probability of generatingan observation sequence?– Model evaluation• Given observation, what is the mostprobable transition sequence?– Segmentation or path analysis• How do we estimate or optimizethe parameters of an HMM?– Training problem0.6 0.61 30.30.20.2 0.60.320.1 0.1P( )?0.6 0.61 30.30.20.2 0.60.320.1 0.12 2 2 21 32?)|,,,( 21 ==λTxxxXP L)|,(maxarg ),,(*1λXQPQ TqqQ L==))',','('|()),,(|(πλπλ BAXPBAXP =<=

How does the state of the process is described in HMM?a) Literalb) Single random variablec) Single discrete random variabled) None of the mentioned

What is the number of parameters in a transition matrix for an HMM with 3 states?Question 8Answera.9b.3c.12d.6

Quiz Question Consider the Markov chain with transition diagram345210.10.50.50.60.30.50.20.40.30.50.50.20.4i.e., with transition matrixP =0.6 0.4 0 0 00.5 0.5 0 0 00 0.5 0 0.5 00 0.5 0.3 0 0.20 0.1 0.2 0.4 0.3 .(a) Starting from state 5, what is the most likely two-step path? What is the probability(starting from state 5) of following that two-step path?(b) In this Markov chain, can you get from every state to every other state eventually?

Which of the following best describes a Hidden Markov Model (HMM)?<br /> A. a. A generative model <br />B. b. A discriminative model <br />C. c. A deterministic model <br />D. d. A rule-based model

Explain HMM model for use for part-of- speech tagging, Chinese word segmentation