Parametric Transfer Learning Based on the Fisher Divergence for Well-Being Prediction

Abstract

Smartphones and wearable sensors are increasingly used for personalised prediction and management in healthcare contexts. Personalisation requires tuning/learning a model of the user. However, traditional machine learning approaches for personalised modelling typically require the availability of sufficient personal data of a suitable nature for training, which can be a challenge in such contexts. We propose a parametric transfer learning approach based on the Fisher divergence to address this challenge. This makes it possible to create patientspecific models and make predictions of self-reported well-being scores, when training is performed incrementally on sparse data becoming slowly available over time. This approach allows us to make informed predictions even in the early stages of data collection, by leveraging external information coming from other patients, in the form of a prior used within a Markov-Chain Monte Carlo process. Our approach performs favourably against competing models and standard baselines, particularly when long-term forecasts are required but training data cover only a short period.

Well-being Forecasting using a Parametric Transfer-Learning method based on the Fisher Divergence and Hamiltonian Monte Carlo

“…This paper is an extended version of the work presented in [22]. We extend our previous work in three important ways: first, by providing a more comprehensive literature review, particularly with respect to the theoretical background underpinning our approach; second, we have included information on diagnostics, detailing how we assessed the quality and convergence of the Markov chains; and third, we present a more rigorous mathematical derivation of the models involved, which was beyond the scope of [22].…”