2019/08/27 by J. Shepard Bryan, Ioannis Sgouralis, Bryan, J Shepard +3 · 2 citations
Biochemistry, Genetics and Molecular Biology · Chemistry · Computer Science · #Biological Physics (physics.bio-ph) #FOS: Physical sciences #Gaussian Processes and Bayesian Inference #Mass Spectrometry Techniques and Applications #Metabolomics and Mass Spectrometry Studies
paper · pdf · doi:10.48550/arxiv.1908.10484
openalex publication_date 2019/08/27 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Effective forces -- derived from experimental or it in silico molecular\ndynamics time traces -- are critical in developing reduced and computationally\nefficient descriptions of otherwise complex dynamical problems. Thus, designing\nmethods to learn effective forces efficiently from time series data is\nimportant. Of equal importance is the fact that methods should be suitable in\ninferring forces for undersampled regions of the phase space where data are\nlimited. Ideally, a method should it a priori be minimally committal as to\nthe shape of the effective force profile, exploit every data point without\nreducing data quality through any form of binning or pre-processing, and\nprovide full credible intervals (error bars) about the prediction. So far no\nmethod satisfies all three criteria. Here we propose a generalization of the\nGaussian process (GP), a key tool in Bayesian nonparametric inference and\nmachine learning, to achieve this for the first time.\n