CS 730/730W/830: Intro AI ■ Break HMMs 1 handout: slides final blog entries were due Wheeler Ruml (UNH) Lecture 27, CS 730 – 1 / 8
Break Wed May 2: HMMs, unsupervised learning, applications ■ ■ Break Mon May 7: special guest Scott Kiesel on robot planning ■ HMMs Wed May 9, 9-noon: project presentations ■ Thur May 10, 8am: paper drafts (optional for some) ■ Fri May 11, 10:30: exam 3 (N133) ■ Tues May 15, 3pm: papers (one hardcopy + electronic PDF) ■ menu? Wheeler Ruml (UNH) Lecture 27, CS 730 – 2 / 8
■ Break HMMs ■ Models ■ The Model ■ Viterbi Decoding ■ Random ■ EOLQs Hidden Markov Models Wheeler Ruml (UNH) Lecture 27, CS 730 – 3 / 8
Probabilistic Models MDPs: ■ Break Naive Bayes: HMMs ■ Models k -Means: ■ The Model Markov chain: ■ Viterbi Decoding ■ Random Hidden Markov model: ■ EOLQs Wheeler Ruml (UNH) Lecture 27, CS 730 – 4 / 8
A Hidden Markov Model ■ Break � P ( x t = j ) = P ( x t − 1 = i ) P ( x t = j | x t − 1 = i ) HMMs ■ Models i ■ The Model � ■ Viterbi Decoding P ( e t = k ) = P ( x t = i ) P ( e = k | x = i ) ■ Random i ■ EOLQs More concisely: � P ( x t ) = P ( x t − 1 ) P ( x t | x t − 1 ) x t − 1 � P ( e t ) = P ( x t ) P ( e | x ) x t Wheeler Ruml (UNH) Lecture 27, CS 730 – 5 / 8
Viterbi Decoding ■ Break given: transition model T ( s, s ′ ) HMMs sensing model S ( s, o ) ■ Models observations o 1 , . . . , o T ■ The Model ■ Viterbi Decoding find: most probable s 1 , . . . , s T ■ Random ■ EOLQs initialize S × T matrix v with 0s v 0 , 0 ← 1 for each time t = 0 to T − 1 for each state s for each new state s ′ score ← v s,t · T ( s, s ′ ) · S ( s ′ , o t ) if score > v s ′ ,t +1 v s ′ ,t +1 ← score best-parent( s ′ ) ← s trace back from s with max v s,T Wheeler Ruml (UNH) Lecture 27, CS 730 – 6 / 8
Random applications ■ Break HMMs ■ Models unsupervised learning: dimensionality reduction ■ The Model ■ Viterbi Decoding ■ Random ■ EOLQs Wheeler Ruml (UNH) Lecture 27, CS 730 – 7 / 8
EOLQs What question didn’t you get to ask today? ■ ■ Break What’s still confusing? ■ HMMs ■ Models What would you like to hear more about? ■ ■ The Model ■ Viterbi Decoding Please write down your most pressing question about AI and put ■ Random ■ EOLQs it in the box on your way out. Thanks! Wheeler Ruml (UNH) Lecture 27, CS 730 – 8 / 8
Recommend
More recommend
