Analysis of the recent article 'Deep learning of aftershock patterns following large earthquakes' by Phoebe M. R. DeVries, Fernanda Viégas, Martin Wattenberg & Brendan J. Meade, published in Nature in 2018.
This repo includes, chronologically:
(1a) notebook folder related to my January 2019 Coursera/IBM Advanced Data Science Capstone project: Aftershock pattern prediction based on earthquake rupture data for improved seismic hazard assessment
(1b) dataset folder and srcmod2.py used in (1a).
(2) ref_Mignan&Broccardo_IWANN2019.pdf, the article A Deeper Look into ‘Deep Learning of Aftershock Patterns Following Large Earthquakes’: Illustrating First Principles in Neural Network Physical Interpretability published in Advances in Computational Intelligence and presented at the International Work-Conference on Artificial Neural Networks (IWANN19) in June 2019. This paper results from (1);
See also:
(3) Coursera/IBM Advanced Data Science Capstone project video
(4) ArXiv preprint One neuron is more informative than a deep neural network for aftershock pattern forecasting where we go one step further in the analysis, showing that a logistic regression (one neuron) performs as well as, if not better than any more complex neural network.
NB: in (1), in case GitHub cannot load a notebook, use instead https://nbviewer.jupyter.org/github/amignan/pred_seism_aftXYZ/blob/master/notebooks/NOTEBOOK_NAME.ipynb
This is for example the case for notebooks containing animations: https://nbviewer.jupyter.org/github/amignan/pred_seism_aftXYZ/blob/master/notebooks/pred_seism_aftXYZ.data_exp.ipynb https://nbviewer.jupyter.org/github/amignan/pred_seism_aftXYZ/blob/master/notebooks/pred_seism_aftXYZ.feature_eng.ipynb