Identifying anomalies in LIGO data by transferring knowledge from artificial intelligence
Researchers from the University of Illinois at Urbana-Champaign and the National Center for Supercomputing Applications Gravity Group expand their novel Deep Filtering method that uses GPU-powered neural networks for anomaly detection and classification of gravitational waves in LIGO data.
“Classification and Clustering of LIGO Data with Deep Transfer Learning”: Accepted for publication in Physical Review D in April 2018
This article shows that deep learning methods can automatically detect and group together anomalies in data from LIGO detectors by using artificial intelligence algorithms based on neural networks that were already pre-trained to classify photographs of real-world objects.
Furthermore, this research shows pre-trained neural networks can be used as feature extractors for unsupervised clustering algorithms to facilitate finding entirely new and unknown classes of glitches and anomalies in gravitational waves without human supervision.
Pre-trained computer vision neural networks were obtained from the Wolfram Neural Net Repository and trained on NVIDIA DGX-1 with Tesla V100 and NVIDIA Tesla P100 GPUs.
“Deep Learning for Real-Time Gravitational Wave Detection and Parameter Estimation: Results with Advanced LIGO Data”: Published in Physics Letters B in March 2018
This article shows for the first time that deep learning can detect true gravitational wave signals in real LIGO data. It is shown that neural networks can be used in realistic detection scenarios and can learn to adapt to the non-Gaussian and non-stationary behavior of real LIGO data. To demonstrate this novel detection method, the researchers showed their method can correctly identify and estimate the properties of gravitational wave detection.
“Deep Neural Networks to Enable Real-Time Multimessenger Astronomy”: Published in Physical Review D in February 2018
The foundation article on deep learning for gravitational wave detection, this paper can be rightly regarded as the textbook reference that established for the first time the power of deep learning to outperform other gravitational wave analysis methods in terms of both accuracy and speed.
This article shows for the very first time that deep convolutional neural networks can match the sensitivity of matched-filtering searches for detecting signals in noisy time series data.