Extending the Reach of Gravitational-wave Detectors with Machine Learning

Document #:

LIGO-T1900359-v1

Document type:

T - Technical notes

Abstract:

This proposal presents the idea of using current machine learning techniques and algorithms to reduce the overall noise floor of the LIGO detectors. There will be a hard emphasis on techniques that analyze time series data, such as utilizing long short-term memory and nonlinear regression algorithms. While other sources of noises in the detectors are outlined in the proposal, there will be a focus on using machine learning algorithms to hone in on noise sources coming from the physical attributes of the instrument itself. The goal is to increase the sensitivity of the detectors by subtracting linear and non-linear noise coupling mechanisms.

Files in Document:

• LIGO Final Report 2019 - Morgan Nanez.pdf (1.0 MB)

Other Files:

• LIGO_Proposal.pdf (93.7 kB)
• LIGO_Proposal_rev.pdf (63.7 kB)
• Final Presentation (SURF 2019) (Nonlinear_Noise_Regression_w_NNs.key, 7.2 MB)
• Interim Report 1 (LIGO_Interim_Report_1.pdf, 471.3 kB)
• Interim Report 2 (before rev) (LIGO_Interim_Report_2_2_0-2.pdf, 1.3 MB)

Topics:

• Data quality / vetoes

Authors:

• Morgan Nanez
• Rana Adhikari
• Michael Coughlin

Keywords:

SURF19

Notes and Changes:

Morgan Nanez SURF19 project material

Referenced by:

• LIGO-T1900757: Final reports from LIGO SURF program, Summer 2019