Objective discovery of dominant dynamical processes with intelligible machine learning
This work addresses the challenge of extracting insights from big data in fields such as climate science and medicine, where existing theory often fails to succinctly describe phenomena, by providing an unsupervised framework for objective discovery.
The authors tackled the problem of discovering dominant dynamical processes in complex natural systems by formulating regime identification as an optimization problem with an intelligible objective function, eliminating the need for a priori knowledge and enabling unbiased data exploration in domains like ocean dynamics and tumor angiogenesis.
The advent of big data has vast potential for discovery in natural phenomena ranging from climate science to medicine, but overwhelming complexity stymies insight. Existing theory is often not able to succinctly describe salient phenomena, and progress has largely relied on ad hoc definitions of dynamical regimes to guide and focus exploration. We present a formal definition in which the identification of dynamical regimes is formulated as an optimization problem, and we propose an intelligible objective function. Furthermore, we propose an unsupervised learning framework which eliminates the need for a priori knowledge and ad hoc definitions; instead, the user need only choose appropriate clustering and dimensionality reduction algorithms, and this choice can be guided using our proposed objective function. We illustrate its applicability with example problems drawn from ocean dynamics, tumor angiogenesis, and turbulent boundary layers. Our method is a step towards unbiased data exploration that allows serendipitous discovery within dynamical systems, with the potential to propel the physical sciences forward.