Senior Project - Pele

Sandia National Laboratories, located in Albuquerque, New Mexico, has developed science-based technologies that support our national security since its inception in 1949. Through science and technology, people, infrastructure, and partnerships, Sandia's mission is to meet national needs in five key areas: Nuclear Weapons, Nonproliferation and Assessments, Military Technologies and Applications, Energy and Infrastructure Assurance, and Homeland Security. In particular, Sandia is interested in automated detection of explosive events in seismic data, specifically differentiating between low-yield nuclear explosions and large chemical explosions, and also between man-made explosions and natural seismic events (e.g. earthquakes and volcanos). Researchers have recently discovered a model of computation involving neural microcircuits (or neural networks) called the Maass Liquid State Machine (LSM), which they feel may provide an effective method of detecting interesting events in seismic data.

The purpose of this project was to provide a robust tool to determine whether the ideas presented in the LSM framework can be used to effectively perform automated analysis of seismic events. To do this, the software provides a method for creating many Liquid State Machine neural networks, each with slightly different parameters, to train the networks, and to simulate them with test data to assess the performance of each neural network. The software also provides a way for users to perform custom translation of seismic data to a format compatible with the LSM framework. The system was written in Java with some C++ components and uses MATLAB as the computational back-end.