Organizers will ensure that any \"meter readings that can be publicly downloaded\" are not included in the private set.\nBut how to define \"meter readings that can be publicly downloaded\" strictly?
\n\nI hope not only EXPLICIT LEAK but also POSSIBLE LEAK will be removed. Otherwise, we have to exploit POSSIBLE LEAK by meaningless reverese engineering or probing in order to improve private LB (I'm afraid internet searching and reverse engineering competitionđ«)
\n\nEXPLICIT LEAK and POSSIBLE LEAK are just my definition. As far as I know,
\n\năEXPLICIT LEAKă\nsite0 / site1(except building152) / site2(except building245)
\n\năPOSSIBLE LEAKă\nbuilding152(site1) / building245(site2) / site3(correlated??) / site4(correlated but not helpful to improve LB?) / site15(at least cor>0.99 part is helpful to imporove LB)
\n\năReferenceă\n0:https://www.kaggle.com/gunesevitan/ashrae-ucf-spider-and-eda-full-test-labels\n1:https://www.kaggle.com/mpware/ucl-data-leakage-episode-2\n2:https://www.kaggle.com/pdnartreb/scrap-asu-data\n4:https://www.kaggle.com/serengil/ucb-data-leakage-site-4-74-buildings\n15:https://www.kaggle.com/pp2file/ashrae-site15-cornell\nThanks you for sincere kagglers.
", "messages": [ { "id": 682318, "postDate": "2019-11-27T08:52:19.080Z", "content": "Organizers will ensure that any \"meter readings that can be publicly downloaded\" are not included in the private set.\nBut how to define \"meter readings that can be publicly downloaded\" strictly?
\n\nI hope not only EXPLICIT LEAK but also POSSIBLE LEAK will be removed. Otherwise, we have to exploit POSSIBLE LEAK by meaningless reverese engineering or probing in order to improve private LB (I'm afraid internet searching and reverse engineering competitionđ«)
\n\nEXPLICIT LEAK and POSSIBLE LEAK are just my definition. As far as I know,
\n\năEXPLICIT LEAKă\nsite0 / site1(except building152) / site2(except building245)
\n\năPOSSIBLE LEAKă\nbuilding152(site1) / building245(site2) / site3(correlated??) / site4(correlated but not helpful to improve LB?) / site15(at least cor>0.99 part is helpful to imporove LB)
\n\năReferenceă\n0:https://www.kaggle.com/gunesevitan/ashrae-ucf-spider-and-eda-full-test-labels\n1:https://www.kaggle.com/mpware/ucl-data-leakage-episode-2\n2:https://www.kaggle.com/pdnartreb/scrap-asu-data\n4:https://www.kaggle.com/serengil/ucb-data-leakage-site-4-74-buildings\n15:https://www.kaggle.com/pp2file/ashrae-site15-cornell\nThanks you for sincere kagglers.
", "rawMarkdown": "Organizers will ensure that any \"meter readings that can be publicly downloaded\" are not included in the private set.\nBut how to define \"meter readings that can be publicly downloaded\" strictly?\n\nI hope not only EXPLICIT LEAK but also POSSIBLE LEAK will be removed. Otherwise, we have to exploit POSSIBLE LEAK by meaningless reverese engineering or probing in order to improve private LB (I'm afraid internet searching and reverse engineering competitionđ«)\n\nEXPLICIT LEAK and POSSIBLE LEAK are just my definition. As far as I know,\n\năEXPLICIT LEAKă\nsite0 / site1(except building152) / site2(except building245)\n\năPOSSIBLE LEAKă\nbuilding152(site1) / building245(site2) / site3(correlated??) / site4(correlated but not helpful to improve LB?) / site15(at least cor>0.99 part is helpful to imporove LB)\n\năReferenceă\n0:https://www.kaggle.com/gunesevitan/ashrae-ucf-spider-and-eda-full-test-labels\n1:https://www.kaggle.com/mpware/ucl-data-leakage-episode-2\n2:https://www.kaggle.com/pdnartreb/scrap-asu-data\n4:https://www.kaggle.com/serengil/ucb-data-leakage-site-4-74-buildings\n15:https://www.kaggle.com/pp2file/ashrae-site15-cornell\nThanks you for sincere kagglers.", "votes": 19 }, { "id": 683429, "postDate": "2019-11-28T12:21:48.557Z", "content": "without a leaked building ids:\nmeter 0: 1413 - 296 = 1117\nmeter 1: 498 - 122 = 376\nmeter 2: 324 - 0 = 324\nmeter 3: 145 - 55 = 90
", "rawMarkdown": "without a leaked building ids:\nmeter 0: 1413 - 296 = 1117\nmeter 1: 498 - 122 = 376\nmeter 2: 324 - 0 = 324\nmeter 3: 145 - 55 = 90", "votes": 1 }, { "id": 682426, "postDate": "2019-11-27T12:28:54.993Z", "content": "I'm afraid that too. It will be the BIG controversial issue after this competitionđ€
", "rawMarkdown": "I'm afraid that too. It will be the **BIG** controversial issue after this competitionđ€ ", "votes": 1, "replies": [ { "id": 682439, "postDate": "2019-11-27T12:39:05.230Z", "content": "Thank you for comment. I respect your success in kernels and public LBđ
", "rawMarkdown": "Thank you for comment. I respect your success in kernels and public LBđ" } ] }, { "id": 682799, "postDate": "2019-11-27T21:44:03.193Z", "content": "I can't really share any concrete information on this subject right now, but I can say that our intent is to err on the side of caution.
", "rawMarkdown": "I can't really share any concrete information on this subject right now, but I can say that our intent is to err on the side of caution.", "votes": 2, "replies": [ { "id": 683422, "postDate": "2019-11-28T12:07:11.420Z", "content": "Thank you. I understand the difficulty. Good to know \"our intent is to err on the side of caution\".
", "rawMarkdown": "Thank you. I understand the difficulty. Good to know \"our intent is to err on the side of caution\"." } ] }, { "id": 682367, "postDate": "2019-11-27T10:42:43.373Z", "content": "how much does lb improve from site 15?
", "rawMarkdown": "how much does lb improve from site 15?", "replies": [ { "id": 682418, "postDate": "2019-11-27T12:18:03.797Z", "content": "0.97(without site15) -> 0.95(with site15)
", "rawMarkdown": "0.97(without site15) -> 0.95(with site15)", "votes": 2 }, { "id": 682421, "postDate": "2019-11-27T12:22:19.567Z", "content": "Wow, ok. Thank you :)
", "rawMarkdown": "Wow, ok. Thank you :)" }, { "id": 682450, "postDate": "2019-11-27T12:53:23.023Z", "content": "@hiroyukiex Thank you so much for information but i got 0.98(without site 4 and 15) and with site 4 and 15 0.97 # can you plz tell me how much change from site 4 if possible(or you are not using site 4 ? )
", "rawMarkdown": "@hiroyukiex Thank you so much for information but i got 0.98(without site 4 and 15) and with site 4 and 15 0.97 # can you plz tell me how much change from site 4 if possible(or you are not using site 4 ? ) ", "votes": 1 }, { "id": 682458, "postDate": "2019-11-27T13:06:30.220Z", "content": "0.95(without site 4) ->0.96(with site 4)\nSite 4 is useless, or I need to look into site4 more deeply (but I do not want to spend time reverse engineering).
", "rawMarkdown": "0.95(without site 4) ->0.96(with site 4)\nSite 4 is useless, or I need to look into site4 more deeply (but I do not want to spend time reverse engineering).", "votes": 4 }, { "id": 682460, "postDate": "2019-11-27T13:12:38.083Z", "content": "Thank you :) i Think i need to check my submission then
", "rawMarkdown": "Thank you :) i Think i need to check my submission then " }, { "id": 682481, "postDate": "2019-11-27T13:47:02.703Z", "content": "That's probably because site 4 has the best score among other sites and site 4 leaks are mess
", "rawMarkdown": "That's probably because site 4 has the best score among other sites and site 4 leaks are mess", "votes": 2 } ] }, { "id": 687918, "postDate": "2019-12-05T01:28:31.933Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 683429, "author_name": "Aleksandr Kosolapov", "author_url": "", "post_date": "2019-11-28T12:21:48.557000", "content": "without a leaked building ids:\nmeter 0: 1413 - 296 = 1117\nmeter 1: 498 - 122 = 376\nmeter 2: 324 - 0 = 324\nmeter 3: 145 - 55 = 90
", "votes": 1, "replies": [] }, { "id": 682426, "author_name": "Isamu", "author_url": "", "post_date": "2019-11-27T12:28:54.993000", "content": "I'm afraid that too. It will be the BIG controversial issue after this competitionđ€
", "votes": 1, "replies": [ { "id": 682439, "author_name": "Hiroyuki Namba", "author_url": "", "post_date": "2019-11-27T12:39:05.230000", "content": "Thank you for comment. I respect your success in kernels and public LBđ
", "votes": 0, "replies": [] } ] }, { "id": 682799, "author_name": "Sohier Dane", "author_url": "", "post_date": "2019-11-27T21:44:03.193000", "content": "I can't really share any concrete information on this subject right now, but I can say that our intent is to err on the side of caution.
", "votes": 2, "replies": [ { "id": 683422, "author_name": "Hiroyuki Namba", "author_url": "", "post_date": "2019-11-28T12:07:11.420000", "content": "Thank you. I understand the difficulty. Good to know \"our intent is to err on the side of caution\".
", "votes": 0, "replies": [] } ] }, { "id": 682367, "author_name": "Oleg Knaub", "author_url": "", "post_date": "2019-11-27T10:42:43.373000", "content": "how much does lb improve from site 15?
", "votes": 0, "replies": [ { "id": 682418, "author_name": "Hiroyuki Namba", "author_url": "", "post_date": "2019-11-27T12:18:03.797000", "content": "0.97(without site15) -> 0.95(with site15)
", "votes": 2, "replies": [] }, { "id": 682421, "author_name": "Oleg Knaub", "author_url": "", "post_date": "2019-11-27T12:22:19.567000", "content": "Wow, ok. Thank you :)
", "votes": 0, "replies": [] }, { "id": 682450, "author_name": "ADITYA KUMAR (Negi)", "author_url": "", "post_date": "2019-11-27T12:53:23.023000", "content": "@hiroyukiex Thank you so much for information but i got 0.98(without site 4 and 15) and with site 4 and 15 0.97 # can you plz tell me how much change from site 4 if possible(or you are not using site 4 ? )
", "votes": 1, "replies": [] }, { "id": 682458, "author_name": "Hiroyuki Namba", "author_url": "", "post_date": "2019-11-27T13:06:30.220000", "content": "0.95(without site 4) ->0.96(with site 4)\nSite 4 is useless, or I need to look into site4 more deeply (but I do not want to spend time reverse engineering).
", "votes": 4, "replies": [] }, { "id": 682460, "author_name": "ADITYA KUMAR (Negi)", "author_url": "", "post_date": "2019-11-27T13:12:38.083000", "content": "Thank you :) i Think i need to check my submission then
", "votes": 0, "replies": [] }, { "id": 682481, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2019-11-27T13:47:02.703000", "content": "That's probably because site 4 has the best score among other sites and site 4 leaks are mess
", "votes": 2, "replies": [] } ] }, { "id": 687918, "author_name": "", "author_url": "", "post_date": "2019-12-05T01:28:31.933000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "682318": "Organizers will ensure that any \"meter readings that can be publicly downloaded\" are not included in the private set.\nBut how to define \"meter readings that can be publicly downloaded\" strictly?\n\nI hope not only EXPLICIT LEAK but also POSSIBLE LEAK will be removed. Otherwise, we have to exploit POSSIBLE LEAK by meaningless reverese engineering or probing in order to improve private LB (I'm afraid internet searching and reverse engineering competitionđ«)\n\nEXPLICIT LEAK and POSSIBLE LEAK are just my definition. As far as I know,\n\năEXPLICIT LEAKă\nsite0 / site1(except building152) / site2(except building245)\n\năPOSSIBLE LEAKă\nbuilding152(site1) / building245(site2) / site3(correlated??) / site4(correlated but not helpful to improve LB?) / site15(at least cor>0.99 part is helpful to imporove LB)\n\năReferenceă\n0:https://www.kaggle.com/gunesevitan/ashrae-ucf-spider-and-eda-full-test-labels\n1:https://www.kaggle.com/mpware/ucl-data-leakage-episode-2\n2:https://www.kaggle.com/pdnartreb/scrap-asu-data\n4:https://www.kaggle.com/serengil/ucb-data-leakage-site-4-74-buildings\n15:https://www.kaggle.com/pp2file/ashrae-site15-cornell\nThanks you for sincere kagglers.", "683429": "without a leaked building ids:\nmeter 0: 1413 - 296 = 1117\nmeter 1: 498 - 122 = 376\nmeter 2: 324 - 0 = 324\nmeter 3: 145 - 55 = 90", "682426": "I'm afraid that too. It will be the **BIG** controversial issue after this competitionđ€ ", "682799": "I can't really share any concrete information on this subject right now, but I can say that our intent is to err on the side of caution.", "682367": "how much does lb improve from site 15?", "687918": "" } }