Dear Kagglers,
\nWe are absolutely delighted by your response to this challenge.
\nThis competition has been quite a rollercoaster, as attested by the evolution
\nof the LB during the competition (specially these last days!). We are aware
\nthe lack of ready-to-go datasets was also a bit of a wall when it came to start
\nthis specific competition, but all those fruitful discussions in the forums have
\nproven once more that the Kaggle community is able to collectively solve
\nall those problems and more. Regardless of your final LB, kudos to all of you!
Congratulations also to the winning teams, and to all of you who were contesting those
\ntop positions during these last days, it's been amazing to follow your progress during these months.
Kaggle kernels update to Python 3.8, we will make our best effort to keep
\nthe frozen branch of PyFstat compatible with
\nPython 3.7 so you can all keep your work and go back to it whenever you feel like.
\nAlso, if you find any issues, please report them at the PyFstat issue tracker.
Once more, thank you very much for taking part in this challenge, we hope
\nyou had as much fun cracking it as we had setting it up!
Rodrigo, Michael, Chris
", "messages": [ { "id": 2085761, "postDate": "2023-01-04T11:37:09.243Z", "content": "Dear Kagglers,
\nWe are absolutely delighted by your response to this challenge.
\nThis competition has been quite a rollercoaster, as attested by the evolution
\nof the LB during the competition (specially these last days!). We are aware
\nthe lack of ready-to-go datasets was also a bit of a wall when it came to start
\nthis specific competition, but all those fruitful discussions in the forums have
\nproven once more that the Kaggle community is able to collectively solve
\nall those problems and more. Regardless of your final LB, kudos to all of you!
Congratulations also to the winning teams, and to all of you who were contesting those
\ntop positions during these last days, it's been amazing to follow your progress during these months.
Kaggle kernels update to Python 3.8, we will make our best effort to keep
\nthe frozen branch of PyFstat compatible with
\nPython 3.7 so you can all keep your work and go back to it whenever you feel like.
\nAlso, if you find any issues, please report them at the PyFstat issue tracker.
Once more, thank you very much for taking part in this challenge, we hope
\nyou had as much fun cracking it as we had setting it up!
Rodrigo, Michael, Chris
", "rawMarkdown": "Dear Kagglers,\n\nWe are absolutely delighted by your response to this challenge.\n\nThis competition has been quite a rollercoaster, as attested by the evolution\nof the LB during the competition (specially these last days!). We are aware\nthe lack of ready-to-go datasets was also a bit of a wall when it came to start\nthis specific competition, but all those fruitful discussions in the forums have\nproven once more that the Kaggle community is able to collectively solve\nall those problems and more. Regardless of your final LB, kudos to all of you!\n\nCongratulations also to the winning teams, and to all of you who were contesting those\ntop positions during these last days, it's been amazing to follow your progress during these months.\n\nKaggle kernels update to Python 3.8, we will make our best effort to keep\nthe frozen branch of [PyFstat](https://github.com/PyFstat/PyFstat) compatible with\nPython 3.7 so you can all keep your work and go back to it whenever you feel like.\nAlso, if you find any issues, please report them [at the PyFstat issue tracker](https://github.com/PyFstat/PyFstat/issues/new/choose).\n\nOnce more, thank you very much for taking part in this challenge, we hope\nyou had as much fun cracking it as we had setting it up!\n\nRodrigo, Michael, Chris", "votes": 15 }, { "id": 2087285, "postDate": "2023-01-05T13:49:15.860Z", "content": "Thank you for organizing this wonderful competition! Helped me exploit the the use of hypothesis testing in signal analysis. The resources offered were indeed useful.
", "rawMarkdown": "Thank you for organizing this wonderful competition! Helped me exploit the the use of hypothesis testing in signal analysis. The resources offered were indeed useful.", "votes": 1 }, { "id": 2085927, "postDate": "2023-01-04T13:25:24.750Z", "content": "Definetely rollercoaster but definetely worthwile :) thank you for organizing!
", "rawMarkdown": "Definetely rollercoaster but definetely worthwile :) thank you for organizing!", "votes": 1 }, { "id": 2086459, "postDate": "2023-01-04T19:35:01.970Z", "content": "Dear Rodrigo, Michael and Chris,
\nThank you very much for organising this competition!
\nIt was a good opportunity for us to learn about gravitational wave detection and to practice our data analysis skills.
\nWe came here for the machine learning, but eventually found – like many people here – that algorithmic methods can still have an edge.
\nThe diversity of solutions is impressive, and it might not have been possible in a notebook-only competition (we wouldn’t have been able to use Julia, nor efficiently implement the stack sliding in Python). So it looks like you made the right choice of format.
\nWe hope that the various proposed solutions will allow you and fellow researchers to push the state of the art in CW detection. We would actually be very interested to know your opinion on the results. Do they suggest some possible improvements to existing methods, or rather confirm that current search methods are still SOTA?
\nAll the best,
\nJean-Loup, Inar and Oleg
Hi @jeanlouptastet , @inartimiryasov , @olegruchayskiy , (also CCing @sakvaua since he may be interested in this answer!)
\nIndeed, the results of this challenge have been rather surprising: This is not an easy problem (signals are not visible at all without extensive preprocessing) and each team tackled that in their own way. As shown by @sakvaua in this other thread, there is quite a diverse pool of solutions to learn something from.
\nNow that everything is done and frozen, we will start processing all this data to understand (and hopefully quantify ;) ) how much was achieved during this challenge. I cannot give a specific timeline, but once the analysis is complete I'll make sure of making it public in another pinned post or similar so anyone interested can have a look.
\nAlso, congrats on that 5th position!
\nCheers,
\nRodrigo
Thank you for organizing this wonderful competition! Helped me exploit the the use of hypothesis testing in signal analysis. The resources offered were indeed useful.
", "votes": 1, "replies": [] }, { "id": 2085927, "author_name": "PiotrKlinke", "author_url": "", "post_date": "2023-01-04T13:25:24.750000", "content": "Definetely rollercoaster but definetely worthwile :) thank you for organizing!
", "votes": 1, "replies": [] }, { "id": 2086459, "author_name": "Jean-Loup Tastet", "author_url": "", "post_date": "2023-01-04T19:35:01.970000", "content": "Dear Rodrigo, Michael and Chris,
\nThank you very much for organising this competition!
\nIt was a good opportunity for us to learn about gravitational wave detection and to practice our data analysis skills.
\nWe came here for the machine learning, but eventually found – like many people here – that algorithmic methods can still have an edge.
\nThe diversity of solutions is impressive, and it might not have been possible in a notebook-only competition (we wouldn’t have been able to use Julia, nor efficiently implement the stack sliding in Python). So it looks like you made the right choice of format.
\nWe hope that the various proposed solutions will allow you and fellow researchers to push the state of the art in CW detection. We would actually be very interested to know your opinion on the results. Do they suggest some possible improvements to existing methods, or rather confirm that current search methods are still SOTA?
\nAll the best,
\nJean-Loup, Inar and Oleg
Hi @jeanlouptastet , @inartimiryasov , @olegruchayskiy , (also CCing @sakvaua since he may be interested in this answer!)
\nIndeed, the results of this challenge have been rather surprising: This is not an easy problem (signals are not visible at all without extensive preprocessing) and each team tackled that in their own way. As shown by @sakvaua in this other thread, there is quite a diverse pool of solutions to learn something from.
\nNow that everything is done and frozen, we will start processing all this data to understand (and hopefully quantify ;) ) how much was achieved during this challenge. I cannot give a specific timeline, but once the analysis is complete I'll make sure of making it public in another pinned post or similar so anyone interested can have a look.
\nAlso, congrats on that 5th position!
\nCheers,
\nRodrigo