Robust Bayesian Transfer Learning between Autoregressive Inference Tasks

0549008 20250123094241.320211202235959.9 85114404289 000853014000021 10.1109/ISSC52156.2021.9467857 Robust Bayesian Transfer Learning between Autoregressive Inference Tasks 6 s. E cav_un_epca*0549104978-1-6654-3429-4Proceedings of the 32nd Irish Signals and Systems Conference (ISSC) 2021PiscatawayIEEE2021 autoregressive (AR) model Bayesian transfer learning data-predictive transfer FPD robust transfer cav_un_auth*0403479 Barber Alec UTIA-B Adaptivní systémy Department of Adaptive Systems AS AS IE Ústav teorie informace a automatizace AV ČR, v. v. i. cav_un_auth*0370768 Quinn Anthony UTIA-B Adaptivní systémy Department of Adaptive Systems AS AS IE Ústav teorie informace a automatizace AV ČR, v. v. i. http://library.utia.cas.cz/separaty/2021/AS/quinn-0549008.pdf GA18-15970S GA ČR CZ cav_un_auth*0362986 Bayesian transfer learning typically relies on a complete stochastic dependence specification between source and target learners. We advocate a solution to the Bayesian transfer learning paradigm which adopts Fully Probabilistic Design (FPD) to search for an optimal choice of distribution constrained by probabilistic source knowledge. Using this optimal decision-making strategy, an algorithm for accepting source knowledge is identified but is found to be effectively insensitive to source uncertainty. Therefore, we propose an adaptation of the FPD framework which results in a robust transfer learning algorithm.Experimental evidence gathered via synthetic data shows enhanced performance when employing both optimal algorithms in a low source data predictor variance regime. In a high source data predictor variance setting, only our adapted FPD-optimal algorithm achieves robustness. cav_un_auth*0418083 Irish Signals and Systems Conference (ISSC 2021) /23./ 20210610 20210611 Athlone IE BD 10000 10100 10102 2022 2 PR RVO:67985556 http://hdl.handle.net/11104/0325124 1 cav_un_auth*0418084 Department of Electronic and Electrical Engineering Trinity College Dublin, the University of Dublin IE S 9467857 2021 85114404289 SCOPUS PUBMED 000853014000021 WOS cav_un_epca*0549104 Proceedings of the 32nd Irish Signals and Systems Conference (ISSC) 2021 IEEE 2021 Piscataway 978-1-6654-3429-4