Deeptime: a Python library for machine learning dynamical models from time series data

Abstract Generation and analysis of time-series data is relevant to many quantitative fields ranging from economics fluid mechanics. In the physical sciences, structures such as metastable coherent sets, slow relaxation processes, collective variables, dominant transition pathways or manifolds channels probability flow can be great importance for understanding characterizing kinetic, thermodynamic mechanistic properties system. Deeptime a general purpose Python library offering various tools estimate dynamical models based on including conventional linear learning methods, Markov state (MSMs), Hidden Models Koopman models, well kernel deep approaches VAMPnets MSMs. The largely compatible with scikit-learn, having range Estimator classes these different but in contrast scikit-learn also provides Model classes, e.g. case an MSM, which provide multitude methods compute interesting thermodynamic, kinetic quantities, free energies, times paths. designed ease use easily maintainable extensible code. this paper we introduce main features structure deeptime software. found under https://deeptime-ml.github.io/ .