1* Whether the 7975 test samples are the entire test set or the public test set, i.e. 24%
\n2*
\n\n\nWe have created a test set comprised of both simulated and actual detector data from LIGO Hanford and LIGO Livingston in which simulated continuous-wave signals were added. For all we know, however, it could be the case that an actual continuous-wave signal is present in the real data and our search methods were not sensitive enough to find it, meaning you could be the first ever to detect a continuous gravitational wave!
\n
https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/357201
\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.
\n3* And how did the hosts add labels to the real samples in the test set? Since they are not sure if it contains a signal. Is it like the samples in the training set that have -1 as the label?
\n4* Are all three -1 samples in the training set real samples? Because the simulated samples already know whether they contain a signal or not when they are synthesised, even if the signal cannot be detected.
", "messages": [ { "id": 2014701, "postDate": "2022-11-02T19:02:53.357Z", "content": "1* Whether the 7975 test samples are the entire test set or the public test set, i.e. 24%
\n2*
\n\n\nWe have created a test set comprised of both simulated and actual detector data from LIGO Hanford and LIGO Livingston in which simulated continuous-wave signals were added. For all we know, however, it could be the case that an actual continuous-wave signal is present in the real data and our search methods were not sensitive enough to find it, meaning you could be the first ever to detect a continuous gravitational wave!
\n
https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/357201
\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.
\n3* And how did the hosts add labels to the real samples in the test set? Since they are not sure if it contains a signal. Is it like the samples in the training set that have -1 as the label?
\n4* Are all three -1 samples in the training set real samples? Because the simulated samples already know whether they contain a signal or not when they are synthesised, even if the signal cannot be detected.
", "rawMarkdown": "1* Whether the 7975 test samples are the entire test set or the public test set, i.e. 24%\n\n2*\n> We have created a test set comprised of both simulated and actual detector data from LIGO Hanford and LIGO Livingston in which simulated continuous-wave signals were added. For all we know, however, it could be the case that an actual continuous-wave signal is present in the real data and our search methods were not sensitive enough to find it, meaning you could be the first ever to detect a continuous gravitational wave!\n\nhttps://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/357201\n\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.\n\n3* And how did the hosts add labels to the real samples in the test set? Since they are not sure if it contains a signal. Is it like the samples in the training set that have -1 as the label?\n\n4* Are all three -1 samples in the training set real samples? Because the simulated samples already know whether they contain a signal or not when they are synthesised, even if the signal cannot be detected.\n", "votes": 3 }, { "id": 2015417, "postDate": "2022-11-03T09:00:24.120Z", "content": "Hi @chenlin1999
\nThanks for taking part in the competition. As for your questions:
\n1) The 7975 test samples are the entire test set of which 24% are used to compute the public leaderboard and the remainder are used for the private leaderboard which will decide the final rankings. I think this is pretty standard for non-code Kaggle competitions.
\n2)
\n\n\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.
\n
In short, yes some of the real samples in the test have had simulated signals added. However, all of the real data has be extensively analysed the LVK Collaboration using various different search techniques and we haven't found anything. So any real signals that could be in the real data will be very weak. Unfortunately, this is unavoidable when using real data since we can't ever be 100% sure there isn't a signal.
\n3) The real data is labelled such that if we added a simulated signal it's 1, if we didn't, then it's 0.
\n4) As for the -1 labels, I suggest you have a look at this post which has more details :) https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/363734
\nHope this helps
", "rawMarkdown": "Hi @chenlin1999\n\nThanks for taking part in the competition. As for your questions:\n\n1) The 7975 test samples are the entire test set of which 24% are used to compute the public leaderboard and the remainder are used for the private leaderboard which will decide the final rankings. I think this is pretty standard for non-code Kaggle competitions.\n\n2) \n> Does this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.\n\nIn short, yes some of the real samples in the test have had simulated signals added. However, all of the real data has be extensively analysed the LVK Collaboration using various different search techniques and we haven't found anything. So any real signals that could be in the real data will be very weak. Unfortunately, this is unavoidable when using real data since we can't ever be 100% sure there isn't a signal.\n\n3) The real data is labelled such that if we added a simulated signal it's 1, if we didn't, then it's 0.\n\n4) As for the -1 labels, I suggest you have a look at this post which has more details :) https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/363734\n\nHope this helps", "votes": 1, "replies": [ { "id": 2015483, "postDate": "2022-11-03T10:05:54.637Z", "content": "Thank you very much for your answer, it was very helpful and great three Easter eggs!
", "rawMarkdown": "Thank you very much for your answer, it was very helpful and great three Easter eggs!" } ] }, { "id": 2014703, "postDate": "2022-11-02T19:05:40.657Z", "content": "@rodrigotenorio @michaeljwill @sakvaua
", "rawMarkdown": "@rodrigotenorio @michaeljwill @sakvaua " } ], "comments": [ { "id": 2015417, "author_name": "Michael J. Williams", "author_url": "", "post_date": "2022-11-03T09:00:24.120000", "content": "Hi @chenlin1999
\nThanks for taking part in the competition. As for your questions:
\n1) The 7975 test samples are the entire test set of which 24% are used to compute the public leaderboard and the remainder are used for the private leaderboard which will decide the final rankings. I think this is pretty standard for non-code Kaggle competitions.
\n2)
\n\n\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.
\n
In short, yes some of the real samples in the test have had simulated signals added. However, all of the real data has be extensively analysed the LVK Collaboration using various different search techniques and we haven't found anything. So any real signals that could be in the real data will be very weak. Unfortunately, this is unavoidable when using real data since we can't ever be 100% sure there isn't a signal.
\n3) The real data is labelled such that if we added a simulated signal it's 1, if we didn't, then it's 0.
\n4) As for the -1 labels, I suggest you have a look at this post which has more details :) https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/363734
\nHope this helps
", "votes": 1, "replies": [ { "id": 2015483, "author_name": "Chen Lin", "author_url": "", "post_date": "2022-11-03T10:05:54.637000", "content": "Thank you very much for your answer, it was very helpful and great three Easter eggs!
", "votes": 0, "replies": [] } ] }, { "id": 2014703, "author_name": "Chen Lin", "author_url": "", "post_date": "2022-11-02T19:05:40.657000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2014701": "1* Whether the 7975 test samples are the entire test set or the public test set, i.e. 24%\n\n2*\n> We have created a test set comprised of both simulated and actual detector data from LIGO Hanford and LIGO Livingston in which simulated continuous-wave signals were added. For all we know, however, it could be the case that an actual continuous-wave signal is present in the real data and our search methods were not sensitive enough to find it, meaning you could be the first ever to detect a continuous gravitational wave!\n\nhttps://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/357201\n\nDoes this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.\n\n3* And how did the hosts add labels to the real samples in the test set? Since they are not sure if it contains a signal. Is it like the samples in the training set that have -1 as the label?\n\n4* Are all three -1 samples in the training set real samples? Because the simulated samples already know whether they contain a signal or not when they are synthesised, even if the signal cannot be detected.\n", "2015417": "Hi @chenlin1999\n\nThanks for taking part in the competition. As for your questions:\n\n1) The 7975 test samples are the entire test set of which 24% are used to compute the public leaderboard and the remainder are used for the private leaderboard which will decide the final rankings. I think this is pretty standard for non-code Kaggle competitions.\n\n2) \n> Does this mean that none of the real samples in the test set had a simulated signal added to them? The simulated signal was only added to the simulated noise? Otherwise, a real sample containing a real signal could have both a real signal and a simulated signal.\n\nIn short, yes some of the real samples in the test have had simulated signals added. However, all of the real data has be extensively analysed the LVK Collaboration using various different search techniques and we haven't found anything. So any real signals that could be in the real data will be very weak. Unfortunately, this is unavoidable when using real data since we can't ever be 100% sure there isn't a signal.\n\n3) The real data is labelled such that if we added a simulated signal it's 1, if we didn't, then it's 0.\n\n4) As for the -1 labels, I suggest you have a look at this post which has more details :) https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves/discussion/363734\n\nHope this helps", "2014703": "@rodrigotenorio @michaeljwill @sakvaua " } }