Transfer sequential Monte Carlo: A framework for Bayesian model transfer
Adam Bretherton · Queensland University of Technology · 2023
Model transfer is the task of using information from source domains to improve inference on a target domain with limited data. In real applications it is unclear when to transfer information, which information to transfer and how to transfer this information. For example, consider modelling an invasive species that requires a time-sensitive solution. Here, waiting for more data can be detrimental, necessitating model transfer from previous studies. We develop a new mathematical and computational framework, implementing transfer for statistical models where it was previously not possible. This new framework permitted an extensive comparison of different approaches for model transfer.