History was made on 14th September 2015 when the first direct observation of Gravitational Waves was made. The gravitational wave originated from the merger of two black holes and was detected by the LIGO gravitational wave detectors in Livingston and Hanford. Since then, Gravitational Waves have been recorded and analyzed to detect and study mergers of Black Holes, Neutron Stars, and other cosmic phenomena. Gravitational Wave Astronomy is an emerging field of science that aims to make use of Gravitational Waves to get a better understanding of the Universe. However, since it is a relatively new field, it is fraught with its own set of difficulties and challenges. The detectors used to detect Gravitational Waves are one of the most sensitive on the planet and are therefore subjected to a lot of noise. This is where the application of Data Science and Machine Learning can prove useful.
G2Net Gravitational Wave Detection Challenge
G2Net is a network of Gravitational Wave, Geophysics, and Machine Learning funded by the European Cooperation in Science and Technology(COST). It aims to make use of Data Science and Machine Learning to develop new algorithms to detect gravitational waves from the background noise. In the G2Net Gravitational Wave Detection Challenge, your aim will be to detect Gravitational Wave signals from the merger of 2 Black Holes.
The training data will contain time-series data of Gravitational Waves from 3 Gravitational Waves Interferometers- LIGO Hanford(in the USA), LIGO Livingston(in the USA), and Virgo(in Italy). The training data contains 560,000 records and each of these records consists of 3 time series(one from each detector mentioned above) of 2 seconds that is sampled at 2048Hz. Each of these records also consists of a target variable- a 0 or a 1. A target variable of 0 means that the time series consists of just noise, whereas a target of 1 means that the time series also contains a gravitational wave recording(a signal) along with noise. The training dataset is balanced equally between the two target categories- noise, and noise+signal.
Your task will be to perform binary classification on this data and predict the probability of a new time series being either noise or signal(along with noise). The solutions will be evaluated by measuring the area under the ROC curve between the predicted probability of a time series and its observed target.
Timelines
The competition began on 30th June, 2021. And while the final submission deadline is on 29th September 2021, the deadline for entry into the competition and team merger is 22nd September 2021. So far, there are already 307 competitors from 289 teams who have made 2,288 entries.
Prizes
While the competition might seem challenging and overwhelming, it is still a competition. So obviously there are prizes. The challenge gives out a total of $15,000 to top 3 placeholders.
The first position holder will be given a total of $6000, while the second and the third will be given $5000 and $4000 respectively.
To be a part of the challenge, you just need to have a Kaggle Account. To know more about this challenge and participate in it, click here.
To know more about G2Net, click here.