![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F9052057%2F10b9128208f4b0e3176af507e20c83ec%2FScreenshot%202023-01-22%202.23.45%20PM.png?generation=1674426259914761&alt=media\")
TrackML Particle Tracking Challenge has some similarities with IceCube competition :
\nhttps://www.kaggle.com/competitions/trackml-particle-identification
", "messages": [ { "id": "2109198", "postDate": "01/21/2023 07:56:54", "content": "TrackML Particle Tracking Challenge has some similarities with IceCube competition :
\nhttps://www.kaggle.com/competitions/trackml-particle-identification
", "rawMarkdown": "TrackML Particle Tracking Challenge has some similarities with IceCube competition :\n\nhttps://www.kaggle.com/competitions/trackml-particle-identification", "votes": null }, { "id": "2111375", "postDate": "01/22/2023 20:56:00", "content": "Wow, that's interesting for us. These two competitions are doing the future research based on the particles around us.
", "rawMarkdown": "Wow, that's interesting for us. These two competitions are doing the future research based on the particles around us.", "votes": null }, { "id": "2111388", "postDate": "01/22/2023 21:11:25", "content": "From the top solution is this competition:
\n'I'm sorry, but I did not use much machine learning (only some logistic regression for candidate pruning), but rather classical mathematical modeling with statistics and 3d geometry. This, combined with the fact that I wrote everything in C++ with no dependencies, made the final code quite fast: about 8 minutes per event per cpu core for my final submission. So I believe my code could be a good starting point for the throughput phase.'
\nI wonder if we will see something similar here. Well, since it is a code competition I guess it would not be so easy to use C++.
\nFrom the comments:
\n'@icecuber, I'm in awe and speechless at the same time after having read your approach and thinking \"what have we been doing in the past 3 months\"? You made us feel really stupid. :p It proved machine learning can't outperform laws of physics. (Science is the king!!) I don't know what people can do in the throughput competition, probably just optimizing your code. :) Your solution is absolutely extraordinary and I wonder if CERN will still bother to explore any DL approaches after having read this post. May I ask if you're a physicist or a mathematician?'
Lol
", "rawMarkdown": "From the top solution is this competition:\n\n'I'm sorry, but I did not use much machine learning (only some logistic regression for candidate pruning), but rather classical mathematical modeling with statistics and 3d geometry. This, combined with the fact that I wrote everything in C++ with no dependencies, made the final code quite fast: about 8 minutes per event per cpu core for my final submission. So I believe my code could be a good starting point for the throughput phase.'\n\nI wonder if we will see something similar here. Well, since it is a code competition I guess it would not be so easy to use C++.\n\nFrom the comments:\n'@icecuber, I'm in awe and speechless at the same time after having read your approach and thinking \"what have we been doing in the past 3 months\"? You made us feel really stupid. :p It proved machine learning can't outperform laws of physics. (Science is the king!!) I don't know what people can do in the throughput competition, probably just optimizing your code. :) Your solution is absolutely extraordinary and I wonder if CERN will still bother to explore any DL approaches after having read this post. May I ask if you're a physicist or a mathematician?'\n\nLol", "votes": null }, { "id": "2111435", "postDate": "01/22/2023 22:24:50", "content": "Shouldn't be any reason you can't use c++. Just upload the binaries to a dataset, or heck you can even just do this:
\n
Good to know…let's hope we won't need to go to such lengths
", "rawMarkdown": "Good to know...let's hope we won't need to go to such lengths", "votes": null }, { "id": "2112215", "postDate": "01/23/2023 13:27:23", "content": "I wonder if the winner (@icecuber) has any relation to IceCube….
\n🤔🤔🤔
", "rawMarkdown": "I wonder if the winner (@icecuber) has any relation to IceCube….\n\n🤔🤔🤔", "votes": null }, { "id": "2112477", "postDate": "01/23/2023 16:45:31", "content": "Off topic, but this comment in the 1st place write-up got me laughing out loud… :D
\n\"It's fun to submit to the leaderboard, even when it isn't strictly strategical considering winning chances.\"
", "rawMarkdown": "Off topic, but this comment in the 1st place write-up got me laughing out loud... :D\n\n\"It's fun to submit to the leaderboard, even when it isn't strictly strategical considering winning chances.\"", "votes": null } ], "comments": [ { "id": 2111375, "author_name": "dinowun", "author_url": "", "post_date": "01/22/2023 20:56:00", "content": "Wow, that's interesting for us. These two competitions are doing the future research based on the particles around us.
", "votes": null, "replies": [] }, { "id": 2111388, "author_name": "shlomoron", "author_url": "", "post_date": "01/22/2023 21:11:25", "content": "From the top solution is this competition:
\n'I'm sorry, but I did not use much machine learning (only some logistic regression for candidate pruning), but rather classical mathematical modeling with statistics and 3d geometry. This, combined with the fact that I wrote everything in C++ with no dependencies, made the final code quite fast: about 8 minutes per event per cpu core for my final submission. So I believe my code could be a good starting point for the throughput phase.'
\nI wonder if we will see something similar here. Well, since it is a code competition I guess it would not be so easy to use C++.
\nFrom the comments:
\n'@icecuber, I'm in awe and speechless at the same time after having read your approach and thinking \"what have we been doing in the past 3 months\"? You made us feel really stupid. :p It proved machine learning can't outperform laws of physics. (Science is the king!!) I don't know what people can do in the throughput competition, probably just optimizing your code. :) Your solution is absolutely extraordinary and I wonder if CERN will still bother to explore any DL approaches after having read this post. May I ask if you're a physicist or a mathematician?'
Lol
", "votes": null, "replies": [ { "id": 2111435, "author_name": "kaggleqrdl", "author_url": "", "post_date": "01/22/2023 22:24:50", "content": "Shouldn't be any reason you can't use c++. Just upload the binaries to a dataset, or heck you can even just do this:
\n
Good to know…let's hope we won't need to go to such lengths
", "votes": null, "replies": [] } ] }, { "id": 2112215, "author_name": "dschettler8845", "author_url": "", "post_date": "01/23/2023 13:27:23", "content": "I wonder if the winner (@icecuber) has any relation to IceCube….
\n🤔🤔🤔
", "votes": null, "replies": [] }, { "id": 2112477, "author_name": "roberthatch", "author_url": "", "post_date": "01/23/2023 16:45:31", "content": "Off topic, but this comment in the 1st place write-up got me laughing out loud… :D
\n\"It's fun to submit to the leaderboard, even when it isn't strictly strategical considering winning chances.\"
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2109198": "TrackML Particle Tracking Challenge has some similarities with IceCube competition :\n\nhttps://www.kaggle.com/competitions/trackml-particle-identification", "2111375": "Wow, that's interesting for us. These two competitions are doing the future research based on the particles around us.", "2111388": "From the top solution is this competition:\n\n'I'm sorry, but I did not use much machine learning (only some logistic regression for candidate pruning), but rather classical mathematical modeling with statistics and 3d geometry. This, combined with the fact that I wrote everything in C++ with no dependencies, made the final code quite fast: about 8 minutes per event per cpu core for my final submission. So I believe my code could be a good starting point for the throughput phase.'\n\nI wonder if we will see something similar here. Well, since it is a code competition I guess it would not be so easy to use C++.\n\nFrom the comments:\n'@icecuber, I'm in awe and speechless at the same time after having read your approach and thinking \"what have we been doing in the past 3 months\"? You made us feel really stupid. :p It proved machine learning can't outperform laws of physics. (Science is the king!!) I don't know what people can do in the throughput competition, probably just optimizing your code. :) Your solution is absolutely extraordinary and I wonder if CERN will still bother to explore any DL approaches after having read this post. May I ask if you're a physicist or a mathematician?'\n\nLol", "2111435": "Shouldn't be any reason you can't use c++. Just upload the binaries to a dataset, or heck you can even just do this:\n\n", "2111464": "Good to know...let's hope we won't need to go to such lengths", "2112215": "I wonder if the winner (@icecuber) has any relation to IceCube….\n\n🤔🤔🤔", "2112477": "Off topic, but this comment in the 1st place write-up got me laughing out loud... :D\n\n\"It's fun to submit to the leaderboard, even when it isn't strictly strategical considering winning chances.\"" }, "source": "meta" }