Název:
Bayesian transfer learning between autoregressive inference tasks
Autoři:
Barber, Alec ; Quinn, Anthony Typ dokumentu: Výzkumné zprávy
Rok:
2020
Jazyk:
eng
Edice: Research Report, svazek: 2389
Abstrakt: Bayesian transfer learning typically relies on a complete stochastic dependence speci cation between source and target learners which allows the opportunity for Bayesian conditioning. We advocate that any requirement for the design or assumption of a full model between target and sources is a restrictive form of transfer learning.
Klíčová slova:
autoregression; food-commodities price prediction; FPD; Fully Probabilistic Design; transfer learning Číslo projektu: GA18-15970S (CEP) Poskytovatel projektu: GA ČR