Séminaire (organisé par l’équipe de recherche DI)

Alessandro Antonucci

Chercheur senior, IDSIA (Istituto Dalle Molle di Studi sull’Intelligenza Artificiale), Lugano, Suisse

Imprecise hidden Markov models and their application to time series classification

Jeudi 5 novembre 2015 en salle GI042

Résumé :

Hidden Markov models (HMMs) are widely used probabilistic models of sequential data. As with other probabilistic models, they require the specification of local conditional probability distributions, whose assessment can be too difficult and error-prone, especially when data are scarce or costly to acquire. The imprecise HMM (iHMM) generalizes HMMs by allowing the quantification to be done by sets of, instead of single, probability distributions. iHMMs have the ability to suspend judgment when there is not enough statistical evidence, and can serve as a sensitivity analysis tool for standard non-stationary HMMs. We show efficient inference algorithms to address standard HMM usage such as the computation of likelihoods and most probable explanations, as well as performing filtering and predictive inference. Experiments show that iHMMs produce more reliable inferences without compromising the computational efficiency. Applications to the classification of human actions are presented. Compared to their classifcal counterpart, iHMMs appear especially effective for an early detection of the actions. The problem of trading off promptness and accuracy in the detection is also discussed.


FR SHIC 3272

Collegium UTC/CNRS