Project Background and Motivation

Earthquake data for the past several years has been collected by the California Integrated Seismic Network (CISN), the Community Seismic Network (CSN) at Caltech and a few other such institutions. Current early warning systems however are statistical and do not make use of artificial intelligence to make the process of detecting earthquakes any quicker; they rely on picking algorithms which although have proven to be very effective, but can be made smarter with the use of predictive intelligence of machine learning . Analysis of the patterns of the past earthquakes can help select out a few patterns that have been seen repeatedly over the years to detect potential future threats. Moreover, they can even further refine current statistical techniques.

Deep Learning on the other hand, is a major upcoming advancement in Machine Learning. It has been ticking off several benchmarks on various datasets, previously set by state of art algorithms. Based over philosophy of One Learning, it seems a good idea to explore its application in seismic event analysis.

For my summer fellowship for 2014, I plan to explore fields of Datascience and Machine Learning, primarily Deep Learning and Unsupervised methods to analyse the seismic data, and work towards discovering new relationships between earthquakes and sensor network. The project will also explore usage of predictive intelligence to better the current parameter set used by CSN.

My research will primarily be based on the field of data-science, and exploring Deep learning along the way. The project will be guided by Prof. Julian Bunn, prinicipal computational scientist at Caltech's Centre for Advance Computing Research and Prof. Mani Chandy, professor of Computer Science.

This is my second summer fellowship, the first one being in summer of 2013. In last SURF, my project explored possibility of usage of Machine Learning to detect earthquakes. 

Check this link for last year's project details and conclusions.