Logo: to the web site of Uppsala University

uu.sePublikasjoner fra Uppsala universitet
Endre søk
ExporteraLink to record
Permanent link

Direct link
BETA

Prosjekt

Prosjekttyp/Bidragsform
Project grant
Tittel [sv]
Probabilistisk modellering av dynamiska system
Tittel [en]
Probabilistic modeling of dynamical systems
Abstract [sv]
The amount of available data is rapidly increasing and to bring out useful information from this data, it has to be modeled and analyzed. A model is a compact and interpretable representation of the data. We will provide new models together with new algorithms, allowing us to automatically learn these models from data generated by the ever-present dynamical systems.We use probabilistic models, allowing us to employ powerful and well-developed probability theory to represent and systematically work with the uncertainty that is inherent in most data. Our new models are capable of expressing uncertainty, not only over parameters, but also over structural aspects of the models, such as segmentations and model orders. These models are based on relatively new combinatorial stochastic processes emerging from the rapidly developing Bayesian nonparametric theory. This opens up for a powerful foundation to model complex dynamical systems, where the structure of the system is changing over time (sometimes referred to as switching or hybrid systems). However, as we seek more flexible models, the inference problems become increasingly more difficult. During the last five years we have witnessed (and contributed to) the development of a very powerful class of inference algorithms, constituted by an elegant mix of particle filtering and Markov chain Monte Carlo methods. This provides us with the mathematical basis we need to tackle the challenging inference problems in this project.
Principal InvestigatorSchön, Thomas
Koordinerande organisasjon
Uppsala universitet
Forskningsfinansiär
Tidsperiod
2014-01-01 - 2017-12-31
HSV kategori
Control EngineeringSignal Processing
Identifikatorer
DiVA, id: project:4984Prosjekt id: 2013-05524_VR

Søk i DiVA

Control EngineeringSignal Processing

Søk utenfor DiVA

GoogleGoogle Scholar