2019 CSDMS meeting-122

From CSDMS





Log in (or create account for non-CSDMS members)
Forgot username? Search or email:CSDMSweb@colorado.edu



Browse  abstracts



Efforts in Machine Learning and Deep Learning to Data-Intensive Problems in Earthquakes and non-linear Geophysical Problems

David Yuen, Columbia University New York New York, United States. daveyuen@gmail.com


The recent incursion of Data Analytics and Big Data has inspired many fields to venture in. Although a late comer, as compared to financial and bioinformatic areas, geosciences have fast picked up momentum in past two years. We will summarize here quantitative efforts, which require computational means beyond a laptop, in machine learning, deep learning and visualization. The examples will be drawn from (1) delineation of three-dimensional sub-surface three -dimensional fault structure illuminated by tens of thousands of hypocenter from earthquake aftershocks in central Italy using unsupervised machine learning (2) Recurrent Neural Networks (RNN) for delineating earthquake Patterns Based on Complete Seismic Catalog created by large-scale finite element Modelling (3) A highly efficient computational interactive Virtual Reality (VR) Visualization Framework and workflow for Geophysical exploration (4) forecasting the intensity trend of the Earth's natural electromagnetic pulse field signal prior to large earthquakes using chaos theory and radial basis functions (RBF) as deep neural network.