Figure - available from: Classical and Quantum Gravity
This content is subject to copyright. Terms and conditions apply.
Comparison of glitches caused by light scattering. (Top) Slow scattering shows up as long duration arches and is correlated with an increase in ground motion in the earthquake (0.03–0.1 Hz) and microseismic (0.1–0.3 Hz) band. (Bottom) Fast scattering often consists of clusters of arches with a higher frequency, and is more common during an increase in anthropogenic band ground motion (1–6 Hz). The noise transients shown here are during the passing of a train close to the Livingston site when fast scattering triggers occur in clusters.
Source publication
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...
Citations
... The classification of triggers is done with Gravity Spy. Gravity Spy is a machine learning project that uses convolutional neural networks to classify transient noise events based on their glitch morphology [14][15][16][17][18]. Other tools, including Hveto, Lasso, iDQ and Detector Characterization Summary pages are used to study the noise correlations between the primary GW channel and different detector components [19][20][21][22]. ...
... Along with the prototypical long duration scattering arches present in earlier observing runs known as Slow Scattering, we noticed the presence of short duration scattering arches in O3. This population was named 'Fast Scattering' due to its high frequency arches [17,28]. In the time-frequency spectrograms, slow scattering arches have a typical duration of 1 s or more, whereas fast scattering arches are much shorter in duration (~0.2 s). ...
... The rate of 4 Hz noise has been found to be correlated with an increase in anthropogenic ground motion. Human activity, trains, bad weather conditions, road work near the site all contribute to an increased rate [17,47]. Figure 4 shows the noise in h(t) strain spectrogram and the 1-5 Hz band ground motion for 21 November 2019. ...
The sensitivity of aLIGO detectors is adversely affected by the presence of noise caused by light scattering. Low frequency seismic disturbances can create higher frequency scattering noise adversely impacting the frequency band in which we detect gravitational waves. In this paper, we analyze instances of a type of scattered light noise we call ‘Fast Scatter’ that is produced by motion at frequencies greater than 1 Hz, to locate surfaces in the detector that may be responsible for the noise. We model the phase noise to better understand the relationship between increases in seismic noise near the site and the resulting Fast Scatter observed. We find that mechanical damping of the arm cavity baffles led to a significant reduction of this noise in recent data. For a similar degree of seismic motion in the 1– 3 Hz range, the rate of noise transients is reduced by a factor of ~50.
... Current observations have characterized dozens of different glitches by Gravity Spy [47,48]. These glitches can exhibit variations in terms of their frequency, amplitude, and duration, making their identification and miti-gation complex tasks. ...
Gravitational wave (GW) detection is of paramount importance in fundamental physics and GW astronomy, yet it presents formidable challenges. One significant challenge is the removal of noise transient artifacts known as ``glitches," which greatly impact the search and identification of GWs. Recent research has achieved remarkable results in data denoising, often using effective modeling methods to remove glitches. However, for glitches from uncertain or unknown sources, current methods cannot completely eliminate them from the GW signal. In this work, we leverage the inherent robustness of machine learning to obtain reliable posterior parameter distributions directly from GW data contaminated by glitches. Our network model provides reasonable and rapid parameter inference even in the presence of glitches, without needing to remove them. We also investigate various factors affecting the rationality of parameter inference in our normalizing flow network, including glitch and GW parameters. The results demonstrate that the normalizing flow can reasonably infer the source parameters of GWs even with unknown contamination. We find that the nature of the glitch itself is the only factor that can affect the rationality of the inferred results. With improvements to our model, we anticipate accelerating the localization of electromagnetic counterparts and providing priors for more accurate deglitching, thereby speeding up subsequent data processing procedures.
... There is a fairly large body of works pertaining to various aspects ranging from denoising, glitch classification and cancellation, waveform modelling, searches for GW signals, astrophysical parameter estimation, population studies (see e.g. [38][39][40][41] for recent reviews). ...
The search for gravitational-wave (GW) signals is limited by non-Gaussian transient noises that mimic astrophysical signals. Temporal coincidence between two or more detectors is used to mitigate contamination by these instrumental glitches. However, when a single detector is in operation, coincidence is impossible, and other strategies have to be used. We explore the possibility of using neural network classifiers and present the results obtained with three types of architectures: convolutional neural network, temporal convolutional network, and inception time. The last two architectures are specifically designed to process time-series data. The classifiers are trained on a month of data from the LIGO Livingston detector during the first observing run (O1) to identify data segments that include the signature of a binary black hole merger. Their performances are assessed and compared. We then apply trained classifiers to the remaining three months of O1 data, focusing specifically on single-detector times. The most promising candidate from our search is 4 January 2016 12:24:17 UTC. Although we are not able to constrain the significance of this event to the level conventionally followed in GW searches, we show that the signal is compatible with the merger of two black holes with masses m1=50.7−8.9+10.4M⊙ and m2=24.4−9.3+20.2M⊙ at the luminosity distance of dL=564−338+812Mpc .
... A common tool used for glitch classification is Gravity Spy [19,20]: a convolutional neural network (CNN) image classifier that distinguishes 21 different glitch classes and 1 Chirp class (consisting of data from hardware injections that emulate the behavior of GWs by displacing the detector's test masses [21]) via time-frequency visualizations, a type of spectrograms called omegascans [22]. A CNN is a type of Deep Learning algorithm designed for image classification consisting of an input layer, an output layer, and several hidden layers that extract useful features from the inputs. ...
... Considering previous investigations on improving Gravity Spy that have shown that its inaccuracies in glitch classification tend to be higher in poorly represented classes in the CNN's training set [30], Jarov et al's study recommends significant changes to Gravity Spy for the purpose of a signal-vs-glitch classifier. First, it recommends augmenting the data of the Chirp class, which is morphologically similar to the typical GW events seen in O3 [20]. However, instead of using hardware injections, it uses GW software simulations that allow the generation of more GW samples, compared to the severely underrepresented Chirp class in the original Gravity Spy training set [19]. ...
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.
... Furthermore, we investigated the effect of noise suppression on the loud noise artifact, known as glitch. We use the Gravity Spy database [64][65][66] to obtain various common types of glitches with an estimated SNR larger than 10 and confidence >0.95. We focus on three categories (Blip, Scattered Light, and Koi Fish) since they are known to be problematic to mimic the response of detectors to an actual GW event [14] and thus limit the overall sensitivity of GW searches [20][21][22]67]. ...
... Glitch is a common occurrence with a rate of ≲1 per minute in the LIGO detectors in O3a [6]. To further investigate the performance of noise suppression on these loud non-Gaussian artifacts, we use the Gravity Spy database [64][65][66], which contains a wide range of glitches. The total number of LIGO glitches considered in this work from the first three observing runs (O1, O2, and O3, where O3 is divided into O3a and O3b) is 15 487, 41 497, 101 614 and 144 958 for O1, O2, O3a, and O3b, respectively. ...
With the advent of gravitational-wave astronomy and the discovery of more compact binary coalescences, data quality improvement techniques are desired to handle the complex and overwhelming noise in gravitational wave (GW) observational data. Though recent machine learning-based studies have shown promising results for data denoising, they are unable to precisely recover both the GW signal amplitude and phase. To address such an issue, we develop a deep neural network centered workflow, WaveFormer, for significant noise suppression and signal recovery on observational data from the Laser Interferometer Gravitational-Wave Observatory (LIGO). The WaveFormer has a science-driven architecture design with hierarchical feature extraction across a broad frequency spectrum. As a result, the overall noise and glitch are decreased by more than one order of magnitude and the signal recovery error is roughly 1% and 7% for the phase and amplitude, respectively. Moreover, on 75 reported binary black hole (BBH) events of LIGO we obtain a significant improvement in inverse false alarm rate. Our work highlights the potential of large neural networks in gravitational wave data analysis and, while primarily demonstrated on LIGO data, its adaptable design indicates promise for broader application within the International Gravitational-Wave Observatories Network (IGWN) in future observational runs.
... Supervised learning needs fixed class definitions that are not exhaustive nor representative of all glitch morphologies, as there could be many possible sub-classes to discover [35]. Furthermore, as GW detectors are improved, novel glitch morphologies could arise [40]. Moreover, generating these labels is an expensive task, since ML methods need a lot of examples for training, and experts must vet the labelling procedure. ...
Gravitational-wave (GW) interferometers are able to detect a change in distance of $\sim$ 1/10,000th the size of a proton. Such sensitivity leads to large rates of non-gaussian, transient bursts of noise, also known as glitches, which hinder the detection and parameter estimation of
short- and long-lived GW signals in the main detector strain. Glitches, come in a wide range of frequency-amplitude-time morphologies and may be caused by environmental or instrumental processes, so a key step towards their mitigation is to understand their population. Current approaches for their identification use supervised models to learn their morphology in the main strain with a fixed set of classes, but do not consider relevant information provided by auxiliary channels that monitor the state of the interferometers. In this work, we present an unsupervised algorithm to find anomalous glitches. Firstly, we encode a subset of auxiliary channels from LIGO Livingston in the fractal dimension, which measures the complexity of the signal. For this aim, we speed up the fractal
dimension calculation to encode $1\,$h of data in $11 \, $s. Secondly, we learn the underlying distribution of the data using an autoencoder with cyclic periodic convolutions. In this way, we learn the underlying distribution of glitches and we uncover
unknown glitch morphologies, and overlaps in time between different glitches
and misclassifications. This led to the discovery of $6.6\%$ anomalies in the input
data. The results of this investigation stress the learnable structure of auxiliary channels encoded in fractal dimension and provide a flexible framework for
glitch discovery.
... Compared to the ∼ 10 2 astrophysical events observed over this timespan [6], the number of glitches that have been uploaded to Gravity Spy is > 10 4 times larger. As of the start of the fourth observing run of the Advanced LIGO, Advanced Virgo and KAGRA detectors, 27 morphological classes [16] including the catch-all None-of-the-Above (NOTA) and No Glitch classes are available for volunteer classification of LIGO glitches, with 23 incorporated into the ML classifier [19]. ...
... Gravity Spy has also had a major impact on the understanding of LIGO noise through the identification of new glitch classes. For example, LIGO scientists identified the Low-Frequency Blip class through investigations of many glitches classified by Gravity Spy as regular Blips; the sub-class of Blips at lower frequencies prompted the creation of a new class in the Gravity Spy project [19]. Gravity Spy volunteers play a key role in this process; they can propose new glitch classes, some of which are eventually incorporated into the ML and classifications. ...
... Another notable example is when Gravity Spy volunteers identified the Crown glitch class, which was contemporaneously identified by LIGO scientists as Fast Scattering [19,38]. These glitches were originally classified into the NOTA class or the Scattered Light class, but further investigation noted they should be their own class, separate from another sub-class of scattered light now known as Slow Scattering. ...
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.
... During O3b, after the reaction chain tracking was employed at LIGO Livingston Observatory (LLO) and LIGO Hanford Observatory (LHO) to diminish the slow scattering transient noise, the average fraction of retracted O3b public alerts dropped from 0.55 to 0.21 [8]. However, reduction of transient noise is only sometimes possible as the noise originates in the complex instrumentation and environment of the detector, and new categories of transient noise may always show up as the sensitivity of the detector improves and new subsystems get added [45][46][47][48]. So, a need for prompt, quantifiable, and automated ability to address noise transients and their impact on public alerts from the LIGO-Virgo-KAGRA network remains. ...
The observation of transient gravitational waves is hindered by the presence of transient noise, colloquially referred to as glitches. These glitches can often be misidentified as gravitational waves by searches for unmodeled transients using the excess-power type of methods and sometimes even excite template waveforms for compact binary coalescences while using matched filter techniques. They thus create a significant background in the searches. This background is more critical in getting identified promptly and efficiently within the context of real-time searches for gravitational-wave transients. Such searches are the ones that have enabled multi-messenger astrophysics with the start of the Advanced LIGO and Advanced Virgo data taking in 2015 and they will continue to enable the field for further discoveries. With this work we propose and demonstrate the use of a signal-based test that quantifies the fidelity of the time-frequency decomposition of the putative signal based on first principles on how astrophysical transients are expected to be registered in the detectors and empirically measuring the instrumental noise. It is based on the Q-transform and a measure of the occupancy of the corresponding time-frequency pixels over select time-frequency volumes; we call it “QoQ”. Our method shows a 40% reduction in the number of retraction of public alerts that were issued by the LIGO-Virgo-KAGRA collaborations during the third observing run with negligible loss in sensitivity. Receiver Operator Characteristic measurements suggest the method can be used in online and offline searches for transients, reducing their background significantly.
... In the past few years, applications of machine learning algorithms have been explored for a variety of tasks in GW data analysis, most notably, CBC searches [34][35][36][37], sensitivity improvements of existing search pipelines [38][39][40][41][42][43], non-linear noise modelling and subtraction [44,45] that is useful for early alerts for CBCs [46], parameter estimation [47,48], glitch classification [49][50][51][52][53][54][55], construction of population models [56][57][58][59][60][61], approximate GW detectability [62][63][64][65] and searches for other interesting signals originating from intermediate-mass black hole (IMBH) mergers [66], strongly-lensed events [67], short gamma-ray bursts [43], continuous waves [68] and unknown astrophysical events [69,70]. ...
... However, in γ metric (as described in equation (2)), these injections are well separated with values 26.5 and 44.9, respectively. training: injections in simulated Gaussian noise (INJ), the same set of injections without noise (CLEANINJ), the same Gaussian noise realisations without injections (GN) and a glitch dataset (GL) obtained from GravitySpy [53,55,81,82]. To ensure near absolute fidelity of the training data, we stick to simulated Gaussian noise. ...
... This forces the model to focus on the uncommon features that correspond to the CBC signal exclusively. To prepare the GL dataset, we obtain the class and GPS time information from Gravity Spy [53,55,81,82], cluster the GPS times within 0.1 s, and choose the data slices such that the glitches are uniformly distributed between 0.25-0.75 s of the slice. ...
The ability of deep learning (DL) approaches to learn generalised signal and noise models, coupled with their fast inference on GPUs, holds great promise for enhancing gravitational-wave (GW) searches in terms of speed, parameter space coverage, and search sensitivity. However, the opaque nature of DL models severely harms their reliability. In this work, we meticulously develop a DL model stage-wise and work towards improving its robustness and reliability. First, we address the problems in maintaining the purity of training data by deriving a new metric that better reflects the visual strength of the "chirp" signal features in the data. Using a reduced, smooth representation obtained through a variational auto-encoder (VAE), we build a classifier to search for compact binary coalescence (CBC) signals. Our tests on real LIGO data show an impressive performance of the model. However, upon probing the robustness of the model through adversarial attacks, its simple failure modes were identified, underlining how such models can still be highly fragile. As a first step towards bringing robustness, we retrain the model in a novel framework involving a generative adversarial network (GAN). Over the course of training, the model learns to eliminate the primary modes of failure identified by the adversaries. Although absolute robustness is practically impossible to achieve, we demonstrate some fundamental improvements earned through such training, like sparseness and reduced degeneracy in the extracted features at different layers inside the model. We show that these gains are achieved at practically zero loss in terms of model performance on real LIGO data before and after GAN training. Through a direct search on ~8.8 days of LIGO data, we recover two significant CBC events from GWTC-2.1 [1], GW190519_153544 and GW190521_074359. We also report the search sensitivity obtained from an injection study.
... Figure 1 shows an example of a Tomte glitch and glitches due to light scattering, along with the other glitch classes that are used in this work. Some glitch classification algorithms, including GravitySpy [27][28][29], glitch-vs-signal classification algorithms, including GSpyNetTree [30], and other methods [31] employ time-frequency spectrograms (a variant of which is called a Qscan or an omegascan) [32] and image classification via a convolutional neural network to distinguish between different glitch and signal classes. These algorithms perform well; however, if a single detector captures an event candidate, time-frequency morphology alone may not be enough to confidently distinguish between a true signal and a detector glitch, or a signal coincident with a detector glitch. ...
The LIGO-Virgo-KAGRA (LVK) network of gravitational-wave (GW) detectors have observed many tens of compact binary mergers to date. Transient, non-Gaussian noise excursions, known as "glitches'', can impact signal detection in various ways. They can imitate true signals as well as reduce the confidence of real signals. In this work, we introduce a novel statistical tool to distinguish astrophysical signals from glitches, using their inferred source parameter posterior distributions as a feature set. By modelling both simulated GW signals and real detector glitches with a gravitational waveform model, we obtain a diverse set of posteriors which are used to train a random forest classifier. We show that random forests can identify differences in the posterior distributions for signals and glitches, aggregating these differences to tell apart signals from common glitch types with high accuracy of over 93%. We conclude with a discussion on the regions of parameter space where the classifier is prone to making misclassifications, and the different ways of implementing this tool into LVK analysis pipelines.
