Speakers
Description
The Laser Interferometer Gravitational Wave-Observatory (LIGO) has accumulated more than 4.5 petabytes (Pb) of data in its quest to detect gravitational waves. Furthermore, it is anticipated that the total data accrued will increase by approximately 0.8 petabytes per year. The processing and analysis of the extensive volume of data from LIGO necessitates a tremendous amount of computational resources and time. Among the most computationally demanding stages are the initial steps of signal extraction and classification, which pose significant challenges. There is a critical need for more efficient detection and classification algorithms that can overcome these current challenges. In this study, we introduce a machine learning methodology utilizing a binary classifier to differentiate and categorize the data. More specifically, we train a convolutional neural network (CNN) using approximately 100,000 simulated time series data samples of gravitational waves encompassing four distinct signal categories: glitches, background noise, binary black hole, and sine-gaussian. To facilitate the recognition and classification of data, our approach involves encoding the time series data into images using Gramian Angular Summation Fields (GASF). By employing this encoding technique, we enable our convolutional neural network (CNN) to identify and categorize the data. Our primary objective is to classify the GASF images into one of the following two groups: noise/glitch signals or transient sine-gaussian/binary black hole signals. Our model achieved a testing accuracy of 97%, demonstrating a high effectiveness when compared to other approaches to classification in the literature. However, there is a need for further investigation into the viability of the Gramian Angular Summation Field approach in converting real LIGO strain data into images and how our approach compares to the current standard, which consists of converting signal data to spectrograms by taking fast Fourier transforms. This exploration aims to evaluate the relative effectiveness of the GASF method and determine its potential for practical implementation.
