Gravity Spy Research Articles

Enhancing the rationale of convolutional neural networks for glitch classification in gravitational wave detectors: a visual explanation

In the pursuit of detecting gravitational waves, ground-based interferometers (e.g. LIGO, Virgo, and KAGRA) face a significant challenge: achieving the extremely high sensitivity required to detect fluctuations at distances significantly smaller than the diameter of an atomic nucleus. Cutting-edge materials and innovative engineering techniques have been employed to enhance the stability and precision of the interferometer apparatus over the years. These efforts are crucial for reducing the noise that masks the subtle gravitational wave signals. Various sources of interference, such as seismic activity, thermal fluctuations, and other environmental factors, contribute to the total noise spectra characteristic of the detector. Therefore, addressing these sources is essential to enhance the interferometer apparatus’s stability and precision. Recent research has emphasised the importance of classifying non-stationary and non-Gaussian glitches, employing sophisticated algorithms and machine learning methods to distinguish genuine gravitational wave signals from instrumental artefacts. The time-frequency-amplitude representation of these transient disturbances exhibits a wide range of new shapes, variability, and features, reflecting the evolution of interferometer technology. In this study, we developed a convolutional neural network model to classify glitches using spectrogram images from the Gravity Spy O1 dataset. We employed score-class activation mapping and the uniform manifold approximation and projection algorithm to visualise and understand the classification decisions made by our model. We assessed the model’s validity and investigated the causes of misclassification from these results.

GSpyNetTree: a signal-vs-glitch classifier for gravitational-wave event candidates

Despite achieving sensitivities capable of detecting the extremely small amplitude of gravitational waves (GWs), LIGO and Virgo detector data contain frequent bursts of non-Gaussian transient noise, commonly known as ‘glitches’. Glitches come in various time-frequency morphologies, and they are particularly challenging when they mimic the form of real GWs. Given the higher expected event rate in the next observing run (O4), LIGO-Virgo GW event candidate validation will require increased levels of automation. Gravity Spy, a machine learning tool that successfully classified common types of LIGO and Virgo glitches in previous observing runs, has the potential to be restructured as a compact binary coalescence (CBC) signal-vs-glitch classifier to accurately distinguish between glitches and GW signals. A CBC signal-vs-glitch classifier used for automation must be robust and compatible with a broad array of background noise, new sources of glitches, and the likely occurrence of overlapping glitches and GWs. We present GSpyNetTree, the Gravity Spy Convolutional Neural Network Decision Tree: a multi-CNN classifier using CNNs in a decision tree sorted via total GW candidate mass tested under these realistic O4-era scenarios.

Utilizing Convolutional Autoencoder for Anomaly Detections in LIGO Spectrogram Data

Massive amounts of data generated from the continuous running of LIGO gravitational-wave detectors comes with a need to search for signals within the data. Gravitational-wave data captured by the detector consist of astronomical events or glitches that last seconds. We present an unsupervised learning using convolutional autoencoder trained on the no-glitch Gravity Spy dataset to do anomaly search on spectrogram data. Reconstruction error is used as the basis and multiple windows are used to improve the model. Results on test data show that the model is capable of detecting signals with significant anomalies such as the Chirp or Koi Fish glitch. Meanwhile, detecting subtle anomalies such as the 1400 Hz Ripples is difficult because its reconstruction error is near the range of noise signals. Validating the result on confirmed gravitational-wave signals shows that the model is capable of gravitational-wave detection.

Gravity Spy: lessons learned and a path forward

The Gravity Spy project aims to uncover the origins of glitches, transient bursts of noise that hamper analysis of gravitational-wave data. By using both the work of citizen-science volunteers and machine learning algorithms, the Gravity Spy project enables reliable classification of glitches. Citizen science and machine learning are intrinsically coupled within the Gravity Spy framework, with machine learning classifications providing a rapid first-pass classification of the dataset and enabling tiered volunteer training, and volunteer-based classifications verifying the machine classifications, bolstering the machine learning training set and identifying new morphological classes of glitches. These classifications are now routinely used in studies characterizing the performance of the LIGO gravitational-wave detectors. Providing the volunteers with a training framework that teaches them to classify a wide range of glitches, as well as additional tools to aid their investigations of interesting glitches, empowers them to make discoveries of new classes of glitches. This demonstrates that, when giving suitable support, volunteers can go beyond simple classification tasks to identify new features in data at a level comparable to domain experts. The Gravity Spy project is now providing volunteers with more complicated data that includes auxiliary monitors of the detector to identify the root cause of glitches.

Convolutional neural networks for the classification of glitches in gravitational-wave data streams

We investigate the use of convolutional neural networks (including the modern ConvNeXt network family) to classify transient noise signals (i.e. glitches) and gravitational waves (GWs) in data from the Advanced LIGO detectors. First, we use models with a supervised learning approach, both trained from scratch using the Gravity Spy dataset and employing transfer learning by fine-tuning pre-trained models in this dataset. Second, we also explore a self-supervised approach, pre-training models with automatically generated pseudo-labels. Our findings are very close to existing results for the same dataset, reaching values for the F1 score of 97.18% (94.15%) for the best supervised (self-supervised) model. We further test the models using actual GW signals from LIGO-Virgo’s O3 run. Although trained using data from previous runs (O1 and O2), the models show good performance, in particular when using transfer learning. We find that transfer learning improves the scores without the need for any training on real signals apart from the less than 50 chirp examples from hardware injections present in the Gravity Spy dataset. This motivates the use of transfer learning not only for glitch classification but also for signal classification.

Glitch subtraction from gravitational wave data using adaptive spline fitting

Transient signals of instrumental and environmental origins (‘glitches’) in gravitational wave data elevate the false alarm rate of searches for astrophysical signals and reduce their sensitivity. Glitches that directly overlap astrophysical signals hinder their detection and worsen parameter estimation errors. As the fraction of data occupied by detectable astrophysical signals will be higher in next generation detectors, such problematic overlaps could become more frequent. These adverse effects of glitches can be mitigated by estimating and subtracting them out from the data, but their unpredictable waveforms and large morphological diversity pose a challenge. Subtraction of glitches using data from auxiliary sensors as predictors works but not for the majority of cases. Thus, there is a need for nonparametric glitch mitigation methods that do not require auxiliary data, work for a large variety of glitches, and have minimal effect on astrophysical signals in the case of overlaps. In order to cope with the high rate of glitches, it is also desirable that such methods be computationally fast. We show that adaptive spline fitting, in which the placement of free knots is optimized to estimate both smooth and non-smooth curves in noisy data, offers a promising approach to satisfying these requirements for broadband short-duration glitches, the type that appear quite frequently. The method is demonstrated on glitches drawn from three distinct classes in the Gravity Spy database as well as on the glitch that overlapped the binary neutron star signal GW170817. The impact of glitch subtraction on the GW170817 signal, or those like it injected into the data, is seen to be negligible.

Data quality up to the third observing run of advanced LIGO: Gravity Spy glitch classifications

Understanding the noise in gravitational-wave detectors is central to detecting and interpreting gravitational-wave signals. Glitches are transient, non-Gaussian noise features that can have a range of environmental and instrumental origins. The Gravity Spy project uses a machine-learning algorithm to classify glitches based upon their time–frequency morphology. The resulting set of classified glitches can be used as input to detector-characterisation investigations of how to mitigate glitches, or data-analysis studies of how to ameliorate the impact of glitches. Here we present the results of the Gravity Spy analysis of data up to the end of the third observing run of advanced laser interferometric gravitational-wave observatory (LIGO). We classify 233981 glitches from LIGO Hanford and 379805 glitches from LIGO Livingston into morphological classes. We find that the distribution of glitches differs between the two LIGO sites. This highlights the potential need for studies of data quality to be individually tailored to each gravitational-wave observatory.

Comparison between t-SNE and cosine similarity for LIGO glitches analysis

The first direct detection of gravitational waves brought not just another proof of Einstein’s theory of general relativity but also different questions about the discovery, and new branches of scientific studies have arisen. The Advanced Laser Interferometer Gravitational-Wave Observatory (LIGO), the experiment that performed such detections, has two observatories, one in Hanford-WA and another in Livingston-LA, and operates as a Michelson–Morley interferometer with 4 km-long arms. Each observatory can measure variations in its arm lengths which are 10 000 times smaller than a proton diameter. Because LIGO has such a high sensitivity to length changes, many noise sources such as environmental effects, instrumental misbehavior, and human activities may also interfere. Studying these local intrusions, which we generically call glitches, remains a big challenge for LIGO Scientific Collaboration since they can mimic gravitational waves, polluting the data and decreasing the statistical significance of a signal. This paper compares two methods of glitch classification for nine classes by using glitchgrams. A glitchgram is constructed using only Omicron triggers and represents an event in the time, frequency, and signal-to-noise ratio space. The first method uses the cosine similarity, and the second uses support vector machine (SVM) from an application of t-distributed stochastic neighbor embedding, an unsupervised machine learning technique. The results from each method are compared with Gravity Spy classifications.

Training Process of Unsupervised Learning Architecture for Gravity Spy Dataset

AbstractTransient noise appearing in the data from gravitational‐wave detectors frequently causes problems, such as instability of the detectors and overlapping or mimicking gravitational‐wave signals. Because transient noise is considered to be associated with the environment and instrument, its classification would help to understand its origin and improve the detector's performance. In a previous study, an architecture for classifying transient noise using a time–frequency 2D image (spectrogram) is proposed, which uses unsupervised deep learning combined with variational autoencoder and invariant information clustering. The proposed unsupervised‐learning architecture is applied to the Gravity Spy dataset, which consists of Advanced Laser Interferometer Gravitational‐Wave Observatory (Advanced LIGO) transient noises with their associated metadata to discuss the potential for online or offline data analysis. In this study, focused on the Gravity Spy dataset, the training process of unsupervised‐learning architecture of the previous study is examined and reported.

On improving the performance of glitch classification for gravitational wave detection by using Generative Adversarial Networks

ABSTRACT Spectrogram classification plays an important role in analysing gravitational wave data. In this paper, we propose a framework to improve the classification performance by using Generative Adversarial Networks (GANs). As substantial efforts and expertise are required to annotate spectrograms, the number of training examples is very limited. However, it is well known that deep networks can perform well only when the sample size of the training set is sufficiently large. Furthermore, the imbalanced sample sizes in different classes can also hamper the performance. In order to tackle these problems, we propose a GAN-based data augmentation framework. While standard data augmentation methods for conventional images cannot be applied on spectrograms, we found that a variant of GANs, ProGAN, is capable of generating high-resolution spectrograms which are consistent with the quality of the high-resolution original images and provide a desirable diversity. We have validated our framework by classifying glitches in the Gravity Spy data set with the GAN-generated spectrograms for training. We show that the proposed method can provide an alternative to transfer learning for the classification of spectrograms using deep networks, i.e. using a high-resolution GAN for data augmentation instead. Furthermore, fluctuations in classification performance with small sample sizes for training and evaluation can be greatly reduced. Using the trained network in our framework, we have also examined the spectrograms with label anomalies in Gravity Spy.

Unsupervised learning architecture for classifying the transient noise of interferometric gravitational-wave detectors

In the data obtained by laser interferometric gravitational wave detectors, transient noise with non-stationary and non-Gaussian features occurs at a high rate. This often results in problems such as detector instability and the hiding and/or imitation of gravitational-wave signals. This transient noise has various characteristics in the time–frequency representation, which is considered to be associated with environmental and instrumental origins. Classification of transient noise can offer clues for exploring its origin and improving the performance of the detector. One approach for accomplishing this is supervised learning. However, in general, supervised learning requires annotation of the training data, and there are issues with ensuring objectivity in the classification and its corresponding new classes. By contrast, unsupervised learning can reduce the annotation work for the training data and ensure objectivity in the classification and its corresponding new classes. In this study, we propose an unsupervised learning architecture for the classification of transient noise that combines a variational autoencoder and invariant information clustering. To evaluate the effectiveness of the proposed architecture, we used the dataset (time–frequency two-dimensional spectrogram images and labels) of the Laser Interferometer Gravitational-wave Observatory (LIGO) first observation run prepared by the Gravity Spy project. The classes provided by our proposed unsupervised learning architecture were consistent with the labels annotated by the Gravity Spy project, which manifests the potential for the existence of unrevealed classes.

Classifying the gravitational waves using the deep learning technique

Gravitational waves are related to the concept of vibration of space-time curvature. When the body of heavy masses lies on the four-dimensional space-time and changes their position with turbulence motion then actually they create a disturbance in the space. The disturbance travels outward from the origin having light velocity is known as gravitational waves. Laser Interferometer Gravitational-Wave Observatory (LIGO) scientific teamwork declared the identification of these waves. In this paper, we review Gravitational waves, Detection of gravitational waves, deep learning for the classification of gravitational waves. We design and develop a deep learning system to classification gravitational waves of the dataset ‘Gravity Spy (Gravitational waves)’ that is made up of the LIGO images. The goals of this research are to gain a piece of reasonable and useful knowledge about Gravitational waves and propose an effective deep learning network system to classify the gravitational waves. The accuracy achieved by our model is 99.34%.

Discovering features in gravitational-wave data through detector characterization, citizen science and machine learning

The observation of gravitational waves is hindered by the presence of transient noise (glitches). We study data from the third observing run of the Advanced LIGO detectors, and identify new glitch classes: fast scattering/crown and low-frequency blips. Using training sets assembled by monitoring of the state of the detector, and by citizen-science volunteers, we update the Gravity Spy machine-learning algorithm for glitch classification. We find that fast scattering/crown, linked to ground motion at the detector sites, is especially prevalent, and identify two subclasses linked to different types of ground motion. Reclassification of data based on the updated model finds that ∼27% of all transient noise at LIGO Livingston belongs to the fast scattering class, while ∼8% belongs to the low-frequency blip class, making them the most frequent and fourth most frequent sources of transient noise at that site. Our results demonstrate both how glitch classification can reveal potential improvements to gravitational-wave detectors, and how, given an appropriate framework, citizen-science volunteers may make discoveries in large data sets.

Transfer learning using Tsallis entropy: An application to Gravity Spy

Recently, transfer learning is applied as an efficient and fast method for object detection and image classification. In this paper, we propose a novel structure for transfer learning based on Tsallis entropy to reduce the loss while classifying images. Also, a comparative analysis is conducted with the traditional cross entropy in transfer learning. The results on different datasets show that transfer learning using Tsallis entropy function has higher accuracy and less loss than the classical method. Finally, the application to Gravity Spy verifies efficiency of the proposed method.

Application of dictionary learning to denoise LIGO’s blip noise transients

Data streams of gravitational-wave detectors are polluted by transient noise features, or "glitches", of instrumental and environmental origin. In this work, we investigate the use of total-variation methods and learned dictionaries to mitigate the effect of those transients in the data. We focus on a specific type of transient, "blip" glitches, as this is the most common type of glitch present in the LIGO detectors and their waveforms are easy to identify. We randomly select 80 blip glitches scattered in the data from advanced LIGO's O1 run, as provided by the citizen-science project Gravity Spy. Our results show that dictionary-learning methods are a valid approach to model and subtract most of the glitch contribution in all cases analyzed, particularly at frequencies below $\sim 1$ kHz. The high-frequency component of the glitch is best removed when a combination of dictionaries with different atom length is employed. As a further example, we apply our approach to the glitch visible in the LIGO-Livingston data around the time of merger of binary neutron star signal GW170817, finding satisfactory results. This paper is the first step in our ongoing program to automatically classify and subtract all families of gravitational-wave glitches employing variational methods.

Shifting forms of Engagement: Volunteer Learning in Online Citizen Science

Peer production projects involve people in many tasks, from editing articles to analyzing datasets. To facilitate mastery of these practices, projects offer a number of learning resources, ranging from project-defined FAQsto individually-oriented search tools and communal discussion boards. However, it is not clear which project resources best support participant learning, overall and at different stages of engagement. We draw onSørensen's framework of forms of presence to distinguish three types of engagement with learning resources:authoritative, agent-centered and communal. We assigned resources from the Gravity Spy citizen-science into these three categories and analyzed trace data recording interactions with resources using a mixed-effects logistic regression with volunteer performance as an outcome variable. The findings suggest that engagement with authoritative resources (e.g., those constructed by project organizers) facilitates performance initially. However, as tasks become more difficult, volunteers seek and benefit from engagement with their own agent-centered resources and community-generated resources. These findings suggest a broader scope for the design of learning resources for peer production

Teaching citizen scientists to categorize glitches using machine learning guided training

Existing literature points to scaffolded training as an effective yet resource-intensive approach to help newcomers learn and stay motivated. Experts need to select relevant learning materials and continuously assess learners' progress. Peer production communities such as Wikipedia and Open Source Software Development projects face the additional problem of turning volunteers into productive participants as soon as possible. To address these challenges, we designed and tested a training regime combining scaffolded instruction and machine learning to select learning materials and gradually introduces new materials to individuals as their competences improve. We evaluated the training regime on 386 participants that contribute to Gravity Spy, an online citizen science project where people are asked to categorize glitches to assist scientists in the search for gravitational waves. Volunteers were assigned to one of two conditions; (1) a machine learning guided training (MLGT) system that continuously assesses volunteers skill level and adjusts the learning materials or (2) an unscaffolded training program where all learning materials were administered at once. Our analysis revealed that volunteers in the MLGT condition were more accurate on the categorization task (an average accuracy of 90% vs. 54%), executed more tasks (an average of 228 vs. 121 tasks), and were retained for a longer period (an average of 2.5 vs. 2 sessions) than volunteers in the unscaffolded training. The results suggest that MLGT is an effective pedagogical approach for training volunteers in categorization tasks and increases volunteers’ motivation.

Classifying the unknown: Discovering novel gravitational-wave detector glitches using similarity learning

The observation of gravitational waves from compact binary coalescences by LIGO and Virgo has begun a new era in astronomy. A critical challenge in making detections is determining whether loud transient features in the data are caused by gravitational waves or by instrumental or environmental sources. The citizen-science project \emph{Gravity Spy} has been demonstrated as an efficient infrastructure for classifying known types of noise transients (glitches) through a combination of data analysis performed by both citizen volunteers and machine learning. We present the next iteration of this project, using similarity indices to empower citizen scientists to create large data sets of unknown transients, which can then be used to facilitate supervised machine-learning characterization. This new evolution aims to alleviate a persistent challenge that plagues both citizen-science and instrumental detector work: the ability to build large samples of relatively rare events. Using two families of transient noise that appeared unexpectedly during LIGO's second observing run (O2), we demonstrate the impact that the similarity indices could have had on finding these new glitch types in the Gravity Spy program.

Classification and unsupervised clustering of LIGO data with Deep Transfer Learning

The exquisite sensitivity of the advanced LIGO detectors has enabled the detection of multiple gravitational wave signals. The sophisticated design of these detectors mitigates the effect of most types of noise. However, advanced LIGO data streams are contaminated by numerous artifacts known as glitches: non-Gaussian noise transients with complex morphologies. Given their high rate of occurrence, glitches can lead to false coincident detections, obscure and even mimic gravitational wave signals. Therefore, successfully characterizing and removing glitches from advanced LIGO data is of utmost importance. Here, we present the first application of Deep Transfer Learning for glitch classification, showing that knowledge from deep learning algorithms trained for real-world object recognition can be transferred for classifying glitches in time-series based on their spectrogram images. Using the Gravity Spy dataset, containing hand-labeled, multi-duration spectrograms obtained from real LIGO data, we demonstrate that this method enables optimal use of very deep convolutional neural networks for classification given small training datasets, significantly reduces the time for training the networks, and achieves state-of-the-art accuracy above 98.8%, with perfect precision-recall on 8 out of 22 classes. Furthermore, new types of glitches can be classified accurately given few labeled examples with this technique. Once trained via transfer learning, we show that the convolutional neural networks can be truncated and used as excellent feature extractors for unsupervised clustering methods to identify new classes based on their morphology, without any labeled examples. Therefore, this provides a new framework for dynamic glitch classification for gravitational wave detectors, which are expected to encounter new types of noise as they undergo gradual improvements to attain design sensitivity.

Folksonomies to Support Coordination and Coordination of Folksonomies

Members of highly-distributed groups in online production communities face challenges in achieving coordinated action. Existing CSCW research highlights the importance of shared language and artifacts when coordinating actions in such settings. To better understand how such shared language and artifacts are, not only a guide for, but also a result of collaborative work we examine the development of folksonomies (i.e., volunteer-generated classification schemes) to support coordinated action. Drawing on structuration theory, we conceptualize a folksonomy as an interpretive schema forming a structure of signification. Our study is set in the context of an online citizen-science project, Gravity Spy, in which volunteers label “glitches” (noise events recorded by a scientific instrument) to identify and name novel classes of glitches. Through a multi-method study combining virtual and trace ethnography, we analyze folksonomies and the work of labelling as mutually constitutive, giving folksonomies a dual role: an emergent folksonomy supports the volunteers in labelling images at the same time that the individual work of labelling images supports the development of a folksonomy. However, our analysis suggests that the lack of supporting norms and authoritative resources (structures of legitimation and domination) undermines the power of the folksonomy and so the ability of volunteers to coordinate their decisions about naming novel glitch classes. These results have implications for system design. If we hope to support the development of emergent folksonomies online production communities need to facilitate 1) tag gardening, a process of consolidating overlapping terms of artifacts; 2) demarcate a clear home for discourses around folksonomy disagreements; 3) highlight clearly when decisions have been reached; and 4) inform others about those decisions.