This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Gravity Spy | |
|---|---|
| Name | Gravity Spy |
| Launch | 2016 |
| Type | Citizen science, machine learning, data classification |
| Partners | LIGO Scientific Collaboration, Zooniverse, University of Oregon, Oak Ridge National Laboratory |
| Status | Active (as of 2020s) |
Gravity Spy
Gravity Spy is a citizen‑science project that combines volunteer classifications with machine‑learning techniques to identify and characterize non‑astrophysical transient noise artifacts in data from the Laser Interferometer Gravitational‑Wave Observatory. The project connects public volunteers with gravitational‑wave data to improve signal quality for searches conducted by international collaborations. It bridges institutions involved in experimental physics, computer science, and public engagement.
Gravity Spy links members of the public to strain‑channel spectrograms collected by the LIGO Hanford Observatory, LIGO Livingston Observatory, and later detectors such as GEO600 and Virgo. Volunteers examine spectrograms and ancillary channels to label glitch classes like "blip", "whistle", and "chirp" that interfere with searches for signals from sources including Binary neutron star merger, Binary black hole, Core-collapse supernova, Continuous gravitational wave, and Pulsar searches. The project integrates tools from the Zooniverse platform, machine‑learning frameworks pioneered in collaborations with groups associated with California Institute of Technology, Massachusetts Institute of Technology, University of Glasgow, and University of Sussex.
The project was developed amid a surge of interest following the historic detection reported by teams from LIGO Scientific Collaboration and Virgo Collaboration in the mid‑2010s. Initial development involved partnerships with Zooniverse and research groups at the University of Oregon and Oak Ridge National Laboratory. Design choices drew on precedent from citizen‑science initiatives such as Galaxy Zoo, Foldit, and SETI@home, and on machine‑learning progress exemplified by competitions like those hosted by Kaggle. Funding and oversight intersected with agencies and institutions including National Science Foundation and academic groups at Caltech and MIT, aligning instrument characterization needs with public engagement goals.
Data consist primarily of time‑frequency visualizations (spectrograms) and auxiliary channel summaries produced by the LIGO Data Grid and data quality pipelines used by the LIGO Scientific Collaboration. Methodology combines human classification with convolutional neural networks and ensemble classifiers trained on volunteer‑labeled datasets. Preprocessing pipelines draw on libraries and toolkits common in collaborations at Lawrence Berkeley National Laboratory, CERN‑adjacent computing efforts, and software developed at University of Wisconsin–Milwaukee and Penn State University. Labels from volunteers feed into supervised learning models, cross‑validated against injections and simulated signals from groups such as NINJA project contributors and waveform models developed at Numerical relativity centers like Cornell University and RIT (Rochester Institute of Technology).
The user interface hosted via Zooniverse provides tutorials, workflows, and discussion boards where volunteers interact with scientists affiliated with LIGO Scientific Collaboration and outreach teams at institutions such as University of Birmingham and University of Minnesota. Volunteers range from hobbyists to students and teachers connected to programs at Smithsonian Institution outreach events and festivals like Pint of Science. Community moderation and training combine contributions from educators associated with National Optical Astronomy Observatory partnerships and software usability studies from groups at University College London.
Outputs include curated glitch catalogs that improved detector characterization for searches yielding detections like those reported by the LIGO Scientific Collaboration and Virgo Collaboration, and enhanced veto strategies used in analyses of events resembling signals from GW150914 and GW170817. Gravity Spy–informed classifiers helped reduce false alarms in matched‑filter pipelines developed at Caltech and MIT, and informed commissioning efforts by teams at LIGO Hanford Observatory and LIGO Livingston Observatory. Publications arising from the project involved collaborations with researchers at Max Planck Institute for Gravitational Physics (Albert Einstein Institute), University of Wisconsin–Milwaukee, and University of Glasgow, and influenced machine‑learning research cited in conferences such as NeurIPS and International Conference on Machine Learning.
Challenges include class imbalance, ambiguous morphologies that overlap with true astrophysical signals like those from Compact binary coalescence, and domain transfer when adapting classifiers between observing runs involving upgrades at Advanced LIGO and detector networks that include KAGRA. Volunteer training requires sustained engagement efforts akin to large outreach programs managed by Smithsonian Institution and educational initiatives at National Science Foundation‑funded centers. Data privacy and embargo policies set by collaborations such as LIGO Scientific Collaboration and computing constraints at facilities like LIGO Data Grid impose operational limits.
The project's legacy includes contributing to best practices in citizen‑science machine‑learning hybrids and inspiring related efforts within gravitational‑wave and broader astrophysics communities, including projects at Zooniverse, successor efforts linking to KAGRA data, and cross‑disciplinary initiatives at institutions like Stanford University and University of Cambridge. It influenced methodology in projects such as Gravity Spy 2‑style proposals, informed glitch‑mitigation processes used by LIGO Scientific Collaboration data analysts, and fed into educational collaborations with museums like the American Museum of Natural History.
Category:Citizen science Category:Gravitational wave astronomy Category:Machine learning in astronomy