Interesting deck about our competition.
\n\n", "messages": [ { "id": "354274", "postDate": "07/09/2018 08:27:20", "content": "Interesting deck about our competition.
\n\n", "rawMarkdown": "Interesting deck about our competition.\n\n[new deck][1]\n\n\n [1]: https://indico.cern.ch/event/686555/contributions/2976579/attachments/1680748/2700965/ICHEP_TrackML_Jul052018.pdf", "votes": null }, { "id": "354305", "postDate": "07/09/2018 09:32:07", "content": "@CPMP Thanks for sharing, you can also follow TrackML's tweets here\nhttps://twitter.com/trackmllhc?lang=en
", "rawMarkdown": "CPMP Thanks for sharing, you can also follow TrackML's tweets here\nhttps://twitter.com/trackmllhc?lang=en", "votes": null }, { "id": "354315", "postDate": "07/09/2018 10:01:24", "content": "Some Kagglers share their ideas and observations, in spite of this knowledge took them many hours of their work and sharing it decreases their personal chance to win. But it moves our work forward and builds the community. It is a pity, that the organizers did not think to inform us (Kagglers) about such important things like (slide 24):
\n\n\n\n\nWe had a few rather harmless bugs/feature in the dataset
\n \n\n
\n- \n
Beam spot σz=5.5 mm instead of 5.5 cm
- \n
Sensor thickness twice too big (but radiation length is OK)
I have lost a pair of days to find that there is something wrong with it.
\n\nEDIT: In my age, losing time is a mortal sin. Taking a break. Thank you all and see you in the other competitions.
", "rawMarkdown": "Some Kagglers share their ideas and observations, in spite of this knowledge took them many hours of their work and sharing it decreases their personal chance to win. But it moves our work forward and builds the community. It is a pity, that the organizers did not think to inform us (Kagglers) about such important things like (slide 24):\n\n> We had a few rather harmless bugs/feature in the dataset\n\n> - Beam spot σz=5.5 mm instead of 5.5 cm\n\n> - Sensor thickness twice too big (but radiation length is OK)\n\nI have lost a pair of days to find that there is something wrong with it.\n\nEDIT: In my age, losing time is a mortal sin. Taking a break. Thank you all and see you in the other competitions.", "votes": null }, { "id": "354320", "postDate": "07/09/2018 10:09:09", "content": "@Grzegorz, is there anything specific about the beam spot and the sensor thickness? What impact did they have on our current (buggy) data?
", "rawMarkdown": "Grzegorz, is there anything specific about the beam spot and the sensor thickness? What impact did they have on our current (buggy) data?", "votes": null }, { "id": "354323", "postDate": "07/09/2018 10:13:07", "content": "Grzegorz,
\n\nI'm documenting the bugs, but, indeed, there is not much reaction, see https://www.kaggle.com/c/trackml-particle-identification/discussion/60182 and https://www.kaggle.com/c/trackml-particle-identification/discussion/57117
\n\nI'm disappointed that the organizers have time to speak about the competition in so many conferences and media yet don't find time to answer questions on this forum.
", "rawMarkdown": "Grzegorz, \n\nI'm documenting the bugs, but, indeed, there is not much reaction, see https://www.kaggle.com/c/trackml-particle-identification/discussion/60182 and https://www.kaggle.com/c/trackml-particle-identification/discussion/57117\n\nI'm disappointed that the organizers have time to speak about the competition in so many conferences and media yet don't find time to answer questions on this forum.", "votes": null }, { "id": "354327", "postDate": "07/09/2018 10:17:46", "content": "Thanks, but they did not provide the link to the slides in their tweets.
", "rawMarkdown": "Thanks, but they did not provide the link to the slides in their tweets.", "votes": null }, { "id": "354337", "postDate": "07/09/2018 10:31:46", "content": "Hi @cpmp, \nAs far as we know, the only bug so far are the rare cases of weird particle trajectories(particle seems to bounce), which we are investigating, and we'll inform you when we have the full picture. The others (beam spot 5mm, larger thickness) are really features, since the dataset is still consistent and realistic. Is there any other bug/feature which you see pending ?\nDavid
", "rawMarkdown": "Hi @cpmp, \nAs far as we know, the only bug so far are the rare cases of weird particle trajectories(particle seems to bounce), which we are investigating, and we'll inform you when we have the full picture. The others (beam spot 5mm, larger thickness) are really features, since the dataset is still consistent and realistic. Is there any other bug/feature which you see pending ?\nDavid", "votes": null }, { "id": "354339", "postDate": "07/09/2018 10:33:31", "content": "David, I provide other form of weird tracks than bouncing in the two posts referenced above.
", "rawMarkdown": "David, I provide other form of weird tracks than bouncing in the two posts referenced above.", "votes": null }, { "id": "354361", "postDate": "07/09/2018 11:36:00", "content": "Thanks, we're investigating these as well
", "rawMarkdown": "Thanks, we're investigating these as well", "votes": null }, { "id": "354374", "postDate": "07/09/2018 12:24:48", "content": "Thanks.
", "rawMarkdown": "Thanks.", "votes": null }, { "id": "354465", "postDate": "07/09/2018 16:11:00", "content": "And now we know that they expect the winning submission to be ~0.9 (crazytrackers @ 0.89...) 😁
", "rawMarkdown": "And now we know that they expect the winning submission to be ~0.9 (crazytrackers @ 0.89...) 😁", "votes": null }, { "id": "354575", "postDate": "07/09/2018 20:16:25", "content": "When I read this I immediately checked the leaderboard because I thought crazytrackers was a new entrant who just got 0.89.
", "rawMarkdown": "When I read this I immediately checked the leaderboard because I thought crazytrackers was a new entrant who just got 0.89.", "votes": null } ], "comments": [ { "id": 354305, "author_name": "nicolefinnie", "author_url": "", "post_date": "07/09/2018 09:32:07", "content": "@CPMP Thanks for sharing, you can also follow TrackML's tweets here\nhttps://twitter.com/trackmllhc?lang=en
", "votes": null, "replies": [ { "id": 354327, "author_name": "cpmpml", "author_url": "", "post_date": "07/09/2018 10:17:46", "content": "Thanks, but they did not provide the link to the slides in their tweets.
", "votes": null, "replies": [] } ] }, { "id": 354315, "author_name": "sionek", "author_url": "", "post_date": "07/09/2018 10:01:24", "content": "Some Kagglers share their ideas and observations, in spite of this knowledge took them many hours of their work and sharing it decreases their personal chance to win. But it moves our work forward and builds the community. It is a pity, that the organizers did not think to inform us (Kagglers) about such important things like (slide 24):
\n\n\n\n\nWe had a few rather harmless bugs/feature in the dataset
\n \n\n
\n- \n
Beam spot σz=5.5 mm instead of 5.5 cm
- \n
Sensor thickness twice too big (but radiation length is OK)
I have lost a pair of days to find that there is something wrong with it.
\n\nEDIT: In my age, losing time is a mortal sin. Taking a break. Thank you all and see you in the other competitions.
", "votes": null, "replies": [ { "id": 354320, "author_name": "nicolefinnie", "author_url": "", "post_date": "07/09/2018 10:09:09", "content": "@Grzegorz, is there anything specific about the beam spot and the sensor thickness? What impact did they have on our current (buggy) data?
", "votes": null, "replies": [] }, { "id": 354323, "author_name": "cpmpml", "author_url": "", "post_date": "07/09/2018 10:13:07", "content": "Grzegorz,
\n\nI'm documenting the bugs, but, indeed, there is not much reaction, see https://www.kaggle.com/c/trackml-particle-identification/discussion/60182 and https://www.kaggle.com/c/trackml-particle-identification/discussion/57117
\n\nI'm disappointed that the organizers have time to speak about the competition in so many conferences and media yet don't find time to answer questions on this forum.
", "votes": null, "replies": [] }, { "id": 354337, "author_name": "droussea", "author_url": "", "post_date": "07/09/2018 10:31:46", "content": "Hi @cpmp, \nAs far as we know, the only bug so far are the rare cases of weird particle trajectories(particle seems to bounce), which we are investigating, and we'll inform you when we have the full picture. The others (beam spot 5mm, larger thickness) are really features, since the dataset is still consistent and realistic. Is there any other bug/feature which you see pending ?\nDavid
", "votes": null, "replies": [] }, { "id": 354339, "author_name": "cpmpml", "author_url": "", "post_date": "07/09/2018 10:33:31", "content": "David, I provide other form of weird tracks than bouncing in the two posts referenced above.
", "votes": null, "replies": [] }, { "id": 354361, "author_name": "droussea", "author_url": "", "post_date": "07/09/2018 11:36:00", "content": "Thanks, we're investigating these as well
", "votes": null, "replies": [] }, { "id": 354374, "author_name": "cpmpml", "author_url": "", "post_date": "07/09/2018 12:24:48", "content": "Thanks.
", "votes": null, "replies": [] } ] }, { "id": 354465, "author_name": "johnhsweeney", "author_url": "", "post_date": "07/09/2018 16:11:00", "content": "And now we know that they expect the winning submission to be ~0.9 (crazytrackers @ 0.89...) 😁
", "votes": null, "replies": [ { "id": 354575, "author_name": "macfarll", "author_url": "", "post_date": "07/09/2018 20:16:25", "content": "When I read this I immediately checked the leaderboard because I thought crazytrackers was a new entrant who just got 0.89.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "354274": "Interesting deck about our competition.\n\n[new deck][1]\n\n\n [1]: https://indico.cern.ch/event/686555/contributions/2976579/attachments/1680748/2700965/ICHEP_TrackML_Jul052018.pdf", "354305": "CPMP Thanks for sharing, you can also follow TrackML's tweets here\nhttps://twitter.com/trackmllhc?lang=en", "354315": "Some Kagglers share their ideas and observations, in spite of this knowledge took them many hours of their work and sharing it decreases their personal chance to win. But it moves our work forward and builds the community. It is a pity, that the organizers did not think to inform us (Kagglers) about such important things like (slide 24):\n\n> We had a few rather harmless bugs/feature in the dataset\n\n> - Beam spot σz=5.5 mm instead of 5.5 cm\n\n> - Sensor thickness twice too big (but radiation length is OK)\n\nI have lost a pair of days to find that there is something wrong with it.\n\nEDIT: In my age, losing time is a mortal sin. Taking a break. Thank you all and see you in the other competitions.", "354320": "Grzegorz, is there anything specific about the beam spot and the sensor thickness? What impact did they have on our current (buggy) data?", "354323": "Grzegorz, \n\nI'm documenting the bugs, but, indeed, there is not much reaction, see https://www.kaggle.com/c/trackml-particle-identification/discussion/60182 and https://www.kaggle.com/c/trackml-particle-identification/discussion/57117\n\nI'm disappointed that the organizers have time to speak about the competition in so many conferences and media yet don't find time to answer questions on this forum.", "354327": "Thanks, but they did not provide the link to the slides in their tweets.", "354337": "Hi @cpmp, \nAs far as we know, the only bug so far are the rare cases of weird particle trajectories(particle seems to bounce), which we are investigating, and we'll inform you when we have the full picture. The others (beam spot 5mm, larger thickness) are really features, since the dataset is still consistent and realistic. Is there any other bug/feature which you see pending ?\nDavid", "354339": "David, I provide other form of weird tracks than bouncing in the two posts referenced above.", "354361": "Thanks, we're investigating these as well", "354374": "Thanks.", "354465": "And now we know that they expect the winning submission to be ~0.9 (crazytrackers @ 0.89...) 😁", "354575": "When I read this I immediately checked the leaderboard because I thought crazytrackers was a new entrant who just got 0.89." }, "source": "meta" }