State sequence prediction in imprecise hidden Markov models Jasper De Bock & Gert de Cooman 27 July 2011
Jasper De Bock & Gert de Cooman
Jasper De Bock & Gert de Cooman
Research group SYSTeMS Jasper De Bock Gert de Cooman
Research group SYSTeMS Jasper De Bock Gert de Cooman Filip Hermans Erik Quaeghebeur Keivan Shariatmadar Arthur Van Camp
State sequence prediction in imprecise hidden Markov models
State sequence prediction in imprecise hidden Markov models The imprecise hidden Markov model
Imprecise hidden Markov model A sequence of hidden state variables X 2 X 3 X 1 O 2 O 1 O 3 S 1 (O 1 |X 1 ) S 2 (O 2 |X 2 ) S 3 (O 3 |X 3 ) A sequence of observable variables Jasper De Bock 8
Imprecise hidden Markov model Q 1 (X 1 ) Q 2 (X 2 |X 1 ) Q 2 (X 3 |X 2 ) A sequence of hidden state variables X 2 X 3 X 1 O 2 O 1 O 3 S 1 (O 1 |X 1 ) S 2 (O 2 |X 2 ) S 3 (O 3 |X 3 ) A sequence of observable variables Jasper De Bock 9
Imprecise hidden Markov model Q 1 (X 1 ) Q 2 (X 2 |X 1 ) Q 2 (X 3 |X 2 ) X 2 X 3 X 1 All local models are coherent lower previsions O 2 O 1 O 3 S 1 (O 1 |X 1 ) S 2 (O 2 |X 2 ) S 3 (O 3 |X 3 ) Jasper De Bock 10
Imprecise hidden Markov model Q 1 (X 1 ) Q 2 (X 2 |X 1 ) Q 2 (X 3 |X 2 ) X 2 X 3 X 1 O 2 O 1 O 3 S 1 (O 1 |X 1 ) S 2 (O 2 |X 2 ) S 3 (O 3 |X 3 ) Jasper De Bock 11
State sequence prediction in imprecise hidden Markov models Epistemic Irrelevance
Epistemic irrelevance X 2 X 3 X 1 O 1 O 2 O 3 Conditional on its mother variable , the non-parent non- descendants of any variable in the tree are epistemically irrelevant to this variable and its descendants Jasper De Bock 13
State sequence prediction in imprecise hidden Markov models Recursive construction of a joint model for the imprecise hidden Markov model
Recursive construction of a joint model
Recursive construction of a joint model • Marginal extension • Independent natural extension
Recursive construction of a joint model • Marginal extension • Independent natural extension
State sequence prediction in imprecise hidden Markov models Conditioning the model on the observations
State sequence prediction in imprecise hidden Markov models Conditioning the model on the observations Generalised Bayes rule: An extension of the Bayes rule to imprecise probabilities
State sequence prediction in imprecise hidden Markov models Maximal state sequences
State sequence prediction in We predict the state sequence by calculating a set of optimal sequences imprecise hidden Notion of optimality: maximality Markov models Strict partial ordening: Maximal state sequences: Maximal state sequences
State sequence prediction in We predict the state sequence by calculating a set of optimal sequences imprecise hidden Notion of optimality: maximality Markov models Strict partial ordening: Maximal state sequences: Maximal state sequences
State sequence prediction in imprecise hidden Markov models EstiHMM: an efficient algorithm to determine the maximal state sequences in an imprecise hidden Markov model
State sequence prediction in imprecise hidden Markov models EstiHMM: an efficient algorithm to determine the maximal state sequences in an imprecise hidden Markov model
EstiHMM: an efficient algorithm to determine the maximal sequences
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality X 2 X 3 O 3 O 2
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality X 1 X 2 X 3 O 3 O 1 O 2
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion • A recursive approach
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion • A recursive approach Complexity
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion • A recursive approach Complexity • Theoretical analysis
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion • A recursive approach Complexity • Theoretical analysis • Linear in the number of Rmaximal sequences
EstiHMM: an efficient algorithm to determine the maximal sequences • Principle of optimality • Deriving an alternative Roptimality criterion • A recursive approach Complexity • Theoretical analysis • Linear in the number of Rmaximal sequences • Empirical confirmation
State sequence prediction in imprecise hidden Markov models A first experiment
A first experiment
A first experiment
A first experiment See you at the poster session!
Recommend
More recommend