I'm sure everyone is impressed y the amazing 0.91+ score achieved by Top Quarks. And they improve almost everyday... I don't know if @outrunner will best what they have.
\n\nAnyway, I' m not not in same league for sure.
", "messages": [ { "id": "368757", "postDate": "08/10/2018 17:40:37", "content": "I'm sure everyone is impressed y the amazing 0.91+ score achieved by Top Quarks. And they improve almost everyday... I don't know if @outrunner will best what they have.
\n\nAnyway, I' m not not in same league for sure.
", "rawMarkdown": "I'm sure everyone is impressed y the amazing 0.91+ score achieved by Top Quarks. And they improve almost everyday... I don't know if @outrunner will best what they have.\n\nAnyway, I' m not not in same league for sure.", "votes": null }, { "id": "368766", "postDate": "08/10/2018 18:12:25", "content": "I definitely sure that Top Quarks have supervised learning under the hood. I bet on U-Net :)
", "rawMarkdown": "I definitely sure that Top Quarks have supervised learning under the hood. I bet on U-Net :)", "votes": null }, { "id": "368787", "postDate": "08/10/2018 19:56:05", "content": "Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase of the challenge (and I'll have lost track and moved on to the next thing by then...)
", "rawMarkdown": "Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase of the challenge (and I'll have lost track and moved on to the next thing by then...)", "votes": null }, { "id": "368908", "postDate": "08/11/2018 08:49:09", "content": "\n\n\nSadly, I'm sure we won't see any code scoring above 0.8 until after the second phase
\n
I'm not so sure. My impression was that the organizers intend to open the accuracy winners' codes at the beginning of the throughput phase so everyone can work on optimizing them for performance. I might have misunderstood their intentions, though.
", "rawMarkdown": "> Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase\n\nI'm not so sure. My impression was that the organizers intend to open the accuracy winners' codes at the beginning of the throughput phase so everyone can work on optimizing them for performance. I might have misunderstood their intentions, though.", "votes": null }, { "id": "368911", "postDate": "08/11/2018 09:04:13", "content": "\n\n\nMy impression was that the organizers intend to open the accuracy winners' codes
\n
It is an interesting thought. Where did you get that impression from?
\n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?
\n\nNote that I assume I won't be one of the prize winners (very likely), and I I assume there is some value in our code (maybe less likely ;) ).
", "rawMarkdown": "> My impression was that the organizers intend to open the accuracy winners' codes\n\n It is an interesting thought. Where did you get that impression from? \n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?\n\nNote that I assume I won't be one of the prize winners (very likely), and I I assume there is some value in our code (maybe less likely ;) ).", "votes": null }, { "id": "368980", "postDate": "08/11/2018 14:01:56", "content": "\n\n\nWhere did you get that impression from?
\n
From David Rousseau's reply: https://www.kaggle.com/c/trackml-particle-identification/discussion/61728#361605
\n\n\n\n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?
\n
Yes, it would. More specifically it will put the winners at a disadvantage relative to the position they have gained in the first phase, while not impacting non-winners. On the other hand, it will likely increase the quality of the outcome of the second phase, which is probably the motivation for doing it this way, if my interpretation is correct at all.
", "rawMarkdown": "> Where did you get that impression from?\n\nFrom David Rousseau's reply: https://www.kaggle.com/c/trackml-particle-identification/discussion/61728#361605\n\n> Won't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?\n\nYes, it would. More specifically it will put the winners at a disadvantage relative to the position they have gained in the first phase, while not impacting non-winners. On the other hand, it will likely increase the quality of the outcome of the second phase, which is probably the motivation for doing it this way, if my interpretation is correct at all.", "votes": null }, { "id": "368986", "postDate": "08/11/2018 14:28:06", "content": "I didn't read it the same way as you. I read it as 'Kaggle rules imply that prize winners code become public\". This is not true, there has been competitions where winner code was not made public at all.
\n\nBut I hope you are right !
", "rawMarkdown": "I didn't read it the same way as you. I read it as 'Kaggle rules imply that prize winners code become public\". This is not true, there has been competitions where winner code was not made public at all.\n\nBut I hope you are right !", "votes": null }, { "id": "369478", "postDate": "08/13/2018 06:25:33", "content": "Wow! 0.92 now.\nI remember when the competition started, 0.3+ was considered a good score. Then scores grew up, but for a long time 0.6-0.7 were highest scores (and it is still a silver medal). And then outrunner came with 0.8.
", "rawMarkdown": "Wow! 0.92 now.\nI remember when the competition started, 0.3+ was considered a good score. Then scores grew up, but for a long time 0.6-0.7 were highest scores (and it is still a silver medal). And then outrunner came with 0.8.", "votes": null }, { "id": "369537", "postDate": "08/13/2018 09:53:03", "content": "I did say that 0.9 would be reached, see https://www.kaggle.com/c/trackml-particle-identification/discussion/58614#341269 ;)
\n\nOf course, I could have been wrong, and getting to 0.2 is absolutely amazing.
", "rawMarkdown": "I did say that 0.9 would be reached, see https://www.kaggle.com/c/trackml-particle-identification/discussion/58614#341269 ;)\n\nOf course, I could have been wrong, and getting to 0.2 is absolutely amazing.", "votes": null }, { "id": "369820", "postDate": "08/13/2018 20:38:40", "content": "A last minute change in the 3rd place, with another impressive score.
", "rawMarkdown": "A last minute change in the 3rd place, with another impressive score.", "votes": null }, { "id": "369823", "postDate": "08/13/2018 20:50:53", "content": "What a tough competition!
", "rawMarkdown": "What a tough competition!", "votes": null }, { "id": "370136", "postDate": "08/14/2018 10:32:43", "content": "@Serguey, All prize competitions are tough ;) Congrats on the result. I don't get why you could not find a team mate BTW.
", "rawMarkdown": "Serguey, All prize competitions are tough ;) Congrats on the result. I don't get why you could not find a team mate BTW.", "votes": null }, { "id": "370163", "postDate": "08/14/2018 11:25:22", "content": "\n\n\nAll prize competitions are tough ;)
\n
Yes, I've already understood that. Even in the 'Whales' competition without prizes I fell down from 5-th place to 14-th in the last week.
\n\n\n\n\nCongrats on the result.
\n
Thanks! I'm the 'competitions expert' now. :)
", "rawMarkdown": "> All prize competitions are tough ;)\n\nYes, I've already understood that. Even in the 'Whales' competition without prizes I fell down from 5-th place to 14-th in the last week.\n\n> Congrats on the result. \n\nThanks! I'm the 'competitions expert' now. :)", "votes": null } ], "comments": [ { "id": 368766, "author_name": "okhlopkov", "author_url": "", "post_date": "08/10/2018 18:12:25", "content": "I definitely sure that Top Quarks have supervised learning under the hood. I bet on U-Net :)
", "votes": null, "replies": [ { "id": 368787, "author_name": "johnhsweeney", "author_url": "", "post_date": "08/10/2018 19:56:05", "content": "Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase of the challenge (and I'll have lost track and moved on to the next thing by then...)
", "votes": null, "replies": [] }, { "id": 368908, "author_name": "edwinst", "author_url": "", "post_date": "08/11/2018 08:49:09", "content": "\n\n\nSadly, I'm sure we won't see any code scoring above 0.8 until after the second phase
\n
I'm not so sure. My impression was that the organizers intend to open the accuracy winners' codes at the beginning of the throughput phase so everyone can work on optimizing them for performance. I might have misunderstood their intentions, though.
", "votes": null, "replies": [] }, { "id": 368911, "author_name": "cpmpml", "author_url": "", "post_date": "08/11/2018 09:04:13", "content": "\n\n\nMy impression was that the organizers intend to open the accuracy winners' codes
\n
It is an interesting thought. Where did you get that impression from?
\n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?
\n\nNote that I assume I won't be one of the prize winners (very likely), and I I assume there is some value in our code (maybe less likely ;) ).
", "votes": null, "replies": [] }, { "id": 368980, "author_name": "edwinst", "author_url": "", "post_date": "08/11/2018 14:01:56", "content": "\n\n\nWhere did you get that impression from?
\n
From David Rousseau's reply: https://www.kaggle.com/c/trackml-particle-identification/discussion/61728#361605
\n\n\n\n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?
\n
Yes, it would. More specifically it will put the winners at a disadvantage relative to the position they have gained in the first phase, while not impacting non-winners. On the other hand, it will likely increase the quality of the outcome of the second phase, which is probably the motivation for doing it this way, if my interpretation is correct at all.
", "votes": null, "replies": [] }, { "id": 368986, "author_name": "cpmpml", "author_url": "", "post_date": "08/11/2018 14:28:06", "content": "I didn't read it the same way as you. I read it as 'Kaggle rules imply that prize winners code become public\". This is not true, there has been competitions where winner code was not made public at all.
\n\nBut I hope you are right !
", "votes": null, "replies": [] } ] }, { "id": 369478, "author_name": "sergeyzlobin", "author_url": "", "post_date": "08/13/2018 06:25:33", "content": "Wow! 0.92 now.\nI remember when the competition started, 0.3+ was considered a good score. Then scores grew up, but for a long time 0.6-0.7 were highest scores (and it is still a silver medal). And then outrunner came with 0.8.
", "votes": null, "replies": [ { "id": 369537, "author_name": "cpmpml", "author_url": "", "post_date": "08/13/2018 09:53:03", "content": "I did say that 0.9 would be reached, see https://www.kaggle.com/c/trackml-particle-identification/discussion/58614#341269 ;)
\n\nOf course, I could have been wrong, and getting to 0.2 is absolutely amazing.
", "votes": null, "replies": [] } ] }, { "id": 369820, "author_name": "yuval6967", "author_url": "", "post_date": "08/13/2018 20:38:40", "content": "A last minute change in the 3rd place, with another impressive score.
", "votes": null, "replies": [ { "id": 369823, "author_name": "sergeyzlobin", "author_url": "", "post_date": "08/13/2018 20:50:53", "content": "What a tough competition!
", "votes": null, "replies": [] }, { "id": 370136, "author_name": "cpmpml", "author_url": "", "post_date": "08/14/2018 10:32:43", "content": "@Serguey, All prize competitions are tough ;) Congrats on the result. I don't get why you could not find a team mate BTW.
", "votes": null, "replies": [] }, { "id": 370163, "author_name": "sergeyzlobin", "author_url": "", "post_date": "08/14/2018 11:25:22", "content": "\n\n\nAll prize competitions are tough ;)
\n
Yes, I've already understood that. Even in the 'Whales' competition without prizes I fell down from 5-th place to 14-th in the last week.
\n\n\n\n\nCongrats on the result.
\n
Thanks! I'm the 'competitions expert' now. :)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "368757": "I'm sure everyone is impressed y the amazing 0.91+ score achieved by Top Quarks. And they improve almost everyday... I don't know if @outrunner will best what they have.\n\nAnyway, I' m not not in same league for sure.", "368766": "I definitely sure that Top Quarks have supervised learning under the hood. I bet on U-Net :)", "368787": "Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase of the challenge (and I'll have lost track and moved on to the next thing by then...)", "368908": "> Sadly, I'm sure we won't see any code scoring above 0.8 until after the second phase\n\nI'm not so sure. My impression was that the organizers intend to open the accuracy winners' codes at the beginning of the throughput phase so everyone can work on optimizing them for performance. I might have misunderstood their intentions, though.", "368911": "> My impression was that the organizers intend to open the accuracy winners' codes\n\n It is an interesting thought. Where did you get that impression from? \n\nWon't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?\n\nNote that I assume I won't be one of the prize winners (very likely), and I I assume there is some value in our code (maybe less likely ;) ).", "368980": "> Where did you get that impression from?\n\nFrom David Rousseau's reply: https://www.kaggle.com/c/trackml-particle-identification/discussion/61728#361605\n\n> Won't it put prize winners of first phase at a disadvantage given we can see their code but they don't see ours?\n\nYes, it would. More specifically it will put the winners at a disadvantage relative to the position they have gained in the first phase, while not impacting non-winners. On the other hand, it will likely increase the quality of the outcome of the second phase, which is probably the motivation for doing it this way, if my interpretation is correct at all.", "368986": "I didn't read it the same way as you. I read it as 'Kaggle rules imply that prize winners code become public\". This is not true, there has been competitions where winner code was not made public at all.\n\nBut I hope you are right !", "369478": "Wow! 0.92 now.\nI remember when the competition started, 0.3+ was considered a good score. Then scores grew up, but for a long time 0.6-0.7 were highest scores (and it is still a silver medal). And then outrunner came with 0.8.", "369537": "I did say that 0.9 would be reached, see https://www.kaggle.com/c/trackml-particle-identification/discussion/58614#341269 ;)\n\nOf course, I could have been wrong, and getting to 0.2 is absolutely amazing.", "369820": "A last minute change in the 3rd place, with another impressive score.", "369823": "What a tough competition!", "370136": "Serguey, All prize competitions are tough ;) Congrats on the result. I don't get why you could not find a team mate BTW.", "370163": "> All prize competitions are tough ;)\n\nYes, I've already understood that. Even in the 'Whales' competition without prizes I fell down from 5-th place to 14-th in the last week.\n\n> Congrats on the result. \n\nThanks! I'm the 'competitions expert' now. :)" }, "source": "meta" }