Show simple item record

dc.contributor.authorDepeweg, Sen
dc.contributor.authorHernández-Lobato, JMen
dc.contributor.authorDoshi-Velez, Fen
dc.contributor.authorUdluft, Sen
dc.date.accessioned2020-08-05T23:30:15Z
dc.date.available2020-08-05T23:30:15Z
dc.date.issued2017-01-01en
dc.identifier.urihttps://www.repository.cam.ac.uk/handle/1810/308814
dc.description.abstractWe present an algorithm for policy search in stochastic dynamical systems using model-based reinforcement learning. The system dynamics are described with Bayesian neural networks (BNNs) that include stochastic input variables. These input variables allow us to capture complex statistical patterns in the transition dynamics (e.g. multi-modality and heteroskedasticity), which are usually missed by alternative modeling approaches. After learning the dynamics, our BNNs are then fed into an algorithm that performs random roll-outs and uses stochastic optimization for policy learning. We train our BNNs by minimizing a-divergences with a = 0.5, which usually produces better results than other techniques such as variational Bayes. We illustrate the performance of our method by solving a challenging problem where model-based approaches usually fail and by obtaining promising results in real-world scenarios including the control of a gas turbine and an industrial benchmark.
dc.rightsAll rights reserved
dc.rights.uri
dc.titleLearning and policy search in stochastic dynamical systems with Bayesian neural networksen
dc.typeConference Object
prism.publicationDate2017en
prism.publicationName5th International Conference on Learning Representations, ICLR 2017 - Conference Track Proceedingsen
dc.identifier.doi10.17863/CAM.55902
rioxxterms.versionVoR
rioxxterms.licenseref.urihttp://www.rioxx.net/licenses/all-rights-reserveden
rioxxterms.licenseref.startdate2017-01-01en
rioxxterms.typeConference Paper/Proceeding/Abstracten


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record