Document Type

Poster

Publication Date

Summer 2025

Abstract

LIGO, the Laser Interferometer Gravitational-Wave Observatory, detects gravitational waves by measuring distortions in spacetime from merging black holes or neutron stars. However, the detector’s extreme sensitivity makes it susceptible to non-astrophysical noise sources, producing loud transient signals known as glitches. Auxiliary sensors, which record environmental and instrumental conditions, provide a rich source of information that can be used to predict and classify glitches. Noise events range in cause and morphological shape, and have been categorized by the Gravity Spy project. These glitches complicate gravitational-wave searches, making it crucial to identify and mitigate them. This summer, I worked on developing GOLIATH, a multi-stage machine learning framework designed to improve glitch classification and prediction. GOLIATH (Glitch Observation using Labeled Inputs and Auxiliary Training Heuristics) preprocesses large glitch type-labeled datasets by integrating Gravity Spy classifications of glitches with auxiliary channel data to train multiple individual artificial neural networks to be “experts” in a specific glitch type. In this poster, I will present the design of the GOLIATH pipeline and preliminary results comparing GOLIATH’s performance against existing MLAs such as TITAN and GIANTS. Future work will involve improving the generalization of the model by modifying training techniques.

Download

Share

COinS
 

Rights Statement

In Copyright - Non-Commercial Use Permitted