Machine Learning and Data Mining for Comprehensive Test Ban Treaty Monitoring

Abstract

The Comprehensive Test Ban Treaty (CTBT) is gaining renewed attention in light of growing worldwide interest in mitigating risks of nuclear weapons proliferation and testing. Since the International Monitoring System (IMS) installed the first suite of sensors in the late 1990's, the IMS network has steadily progressed, providing valuable support for event diagnostics. This progress was highlighted at the recent International Scientific Studies (ISS) Conference in Vienna in June 2009, where scientists and domain experts met with policy makers to assess the current status of the CTBT Verification System.A strategic theme within the ISS Conference centered on exploring opportunities for further enhancing the detection and localization accuracy of low magnitude events by drawing upon modern tools and techniques for machine learning and large-scale data analysis. Several promising approaches for data exploitation were presented at the Conference. These are summarized in a companion report. In this paper, we introduce essential concepts in machine learning and assess techniques which could provide both incremental and comprehensive value for event discrimination by increasing the accuracy of the final data product, refining On-Site-Inspection (OSI) conclusions, and potentially reducing the cost of future network operations.

Global Seismic Monitoring: A Bayesian Approach

“…Whereas pipelined approaches using local decisions and hard thresholds are ubiquitous in large sensing systems of all kinds, we believe that a vertically integrated probability model connecting raw data to high-level hypotheses can be far more effective, since it enables all available information to be brought to bear on the interpretation of each locally ambiguous datum. Our work to date (Arora et al 2010;Russell, Vaidya, and Le Bras 2010) demonstrates the effectiveness of even a partial realization of this approach. Our initial system, NET-VISA, handles the network processing stage, 1 relying on the IDC's pre-existing signal detection algorithms.…”

Interactive analysis of the results of NET-VISA, a Bayesian inference system, in CTBTO’s International Data Centre bulletin production

“…While effective in many cases, GA struggles with accuracy in associating events, especially under varying geological and seismic conditions. Addressing this, the Network Processing Vertically Integrated Seismic Analysis (NET-VISA) approach, a Bayesian seismic monitoring system designed to process the IMS data (Sereno and Patnaik 1993;Coyne et al 2009;Russell et al 2010) to reduce the number of missed and false events in the automatic processing, as introduced by Arora et al 2013.…”

How Can Science Support a Process Towards a World Free of Nuclear Weapons?

“…26 As discussed previously, the very first applications to CTBT verification proved most interesting and promising. 27,28 In addition to progress in individual verification technologies and data analysis, science can provide support at a broader system level. System analysis is essential to explore how different components can be combined to predict outcomes.…”