Skip to main content
. Author manuscript; available in PMC: 2019 Jan 9.
Published in final edited form as: Metabolism. 2018 Aug 8;87:A1–A9. doi: 10.1016/j.metabol.2018.08.002

Fig. 2.

Fig. 2.

Multi-fidelity data integration through active leaning: Active learning is combined with the NARGP algorithm to integrate histology, omics and clinical data into a machine-learning predictor. It can also guide us as to what new experiments are needed to enhance predictability, which works by considering the maximum of an acquisition function and obtaining one more point in the parameter space using data at different levels (e.g. low- or high-fidelity – (LF) or (HF)).