Tutorial material for MLINT, an initiative at the NOAA Geophysical Fluid Dynamics Laboratory. We explore the ideas of uncovering dynamical regimes and theoretical discovery using machine learning, using both synthetic data, and data from Sonnewald et al. 2019 & 2021. The material is in the shape of a `binder', so no installation is necessary to execute and play with the code. Simply click the button:
The ``notebook'' uses the python programming language. The different bits are organized into cells that can be executed using shift+enter The hope here is to provide both some basic rationales and code with which to start off any exploration, and I use a mix of real and synthetic toy data towards this.
Maike Sonnewald and Redouane Lguensat. Revealing the impact of global heating on north atlantic circulation using transparent machine learning. JAMES, 2021, https://doi.org/10.1029/2021MS002496
Maike Sonnewald, Carl Wunsch, and Patrick Heimbach. Unsupervised learning reveals geography of global ocean dynamical regions. Earth and Space Science, 6(5):784–794, 2019.
Further reading also about objective assessment metrics can be found here: Kaiser et al., Objective discovery of dominant dynamical processes with intelligible machine learning, 2021: https://arxiv.org/abs/2106.12963