![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113389%2F425347e07b7fcda691ce5e73ee01ba7e%2Fblending.jpg?generation=1573121856673590&alt=media\")
24h, one blending kernel => 90 spots collapse. I've got a feeling it won't be the last :-/
", "messages": [ { "id": "667486", "postDate": "11/07/2019 09:37:26", "content": "24h, one blending kernel => 90 spots collapse. I've got a feeling it won't be the last :-/
", "rawMarkdown": "24h, one blending kernel => 90 spots collapse. I've got a feeling it won't be the last :-/", "votes": null }, { "id": "667515", "postDate": "11/07/2019 10:17:46", "content": "
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F413189%2F9ae1a83da45ef21221ab59a645e4b377%2FBlending.jpg?generation=1573123626992168&alt=media\")
Its the beginning of blending Season. It will be on till competition gets over.
", "rawMarkdown": "Its the beginning of blending Season. It will be on till competition gets over.", "votes": null }, { "id": "667574", "postDate": "11/07/2019 12:00:07", "content": "Be more optimistic - that means different approaches could provide different results. And after you will add something unique to them (like features or processing) you will be able to increase your score at ease. Just don't spend any submission for blending right now - the blending guys did this for you.
", "rawMarkdown": "Be more optimistic - that means different approaches could provide different results. And after you will add something unique to them (like features or processing) you will be able to increase your score at ease. Just don't spend any submission for blending right now - the blending guys did this for you.", "votes": null }, { "id": "667621", "postDate": "11/07/2019 13:11:32", "content": "Sorry mate! There is always a first one ;)\nBut it works as tutorial as well, for people that still don't know the power of the ensembles...
", "rawMarkdown": "Sorry mate! There is always a first one ;)\nBut it works as tutorial as well, for people that still don't know the power of the ensembles...", "votes": null }, { "id": "667633", "postDate": "11/07/2019 13:30:35", "content": "\n\n\nBlending season is upon us
\n
...and it is actually making me really happy. When some get lured in the easy (but probably overfitted) path of LB climbing, others stick to their trusted models. It worked for me in IEEE Fraud Detection at least! :) \n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F539648%2F9e27adb5c45370ccd2bc92f512a265a9%2FCapture.PNG?generation=1573133421901949&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F89435%2F6118cb32e946b197156f8ce07ab48ec5%2FScreenshot_20191107_213318.png?generation=1573155276704829&alt=media\")
I've got a feeling it won't be the last :-/
\n\nwhat you mean ?
", "rawMarkdown": "*I've got a feeling it won't be the last :-/*\n\nwhat you mean ?", "votes": null }, { "id": "667907", "postDate": "11/07/2019 20:08:53", "content": "", "rawMarkdown": "", "votes": null }, { "id": "667912", "postDate": "11/07/2019 20:10:25", "content": "But it works as tutorial as well
\n\nare you really sure ?
", "rawMarkdown": "*But it works as tutorial as well*\n\nare you really sure ?", "votes": null }, { "id": "667917", "postDate": "11/07/2019 20:16:51", "content": "So what ? (Miles Davis)
\n\nhttps://www.youtube.com/watch?v=zqNTltOGh5c
\n\nlisten to that and keep cool 🤘
\n\nSure you will recover once.
", "rawMarkdown": "So what ? (Miles Davis) \n\nhttps://www.youtube.com/watch?v=zqNTltOGh5c\n\nlisten to that and keep cool 🤘 \n\nSure you will recover once.", "votes": null }, { "id": "667918", "postDate": "11/07/2019 20:17:40", "content": "Good atmosphere here !
\n\nThanks, I will be back ...
", "rawMarkdown": "Good atmosphere here ! \n\nThanks, I will be back ...", "votes": null }, { "id": "667920", "postDate": "11/07/2019 20:18:55", "content": "Heewww ! It is so cold where you live.... Brrr Brrr Brr.
\n\nWait for tomorrow. Sun will be there, for sure !
", "rawMarkdown": "Heewww ! It is so cold where you live.... Brrr Brrr Brr.\n\nWait for tomorrow. Sun will be there, for sure !", "votes": null }, { "id": "667926", "postDate": "11/07/2019 20:25:15", "content": "It's all over now... Baby Blue 🤜
\n\nhttps://www.youtube.com/watch?v=af7ngGxEusE
", "rawMarkdown": "It's all over now... Baby Blue 🤜 \n\nhttps://www.youtube.com/watch?v=af7ngGxEusE", "votes": null }, { "id": "667952", "postDate": "11/07/2019 20:54:28", "content": "\n\nAllways been a follower and a supporter, really !
", "rawMarkdown": "konradb \n\nAllways been a follower and a supporter, really !", "votes": null }, { "id": "667954", "postDate": "11/07/2019 20:59:40", "content": "Hem ... Explain me what a blend can add to one competition, except but blending... Heeww (Homer Simpson saying).
", "rawMarkdown": "Hem ... Explain me what a blend can add to one competition, except but blending... Heeww (Homer Simpson saying).", "votes": null }, { "id": "667958", "postDate": "11/07/2019 21:03:59", "content": "That's *some *footage btw!
", "rawMarkdown": "That's *some *footage btw!", "votes": null }, { "id": "667973", "postDate": "11/07/2019 21:12:46", "content": "\n\nYou know what ? This competition is really oughta sight ! That's a weird competition :
\n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/114893#latest-666521
\n\nWhat a feeling having so much DownVotes...
\n\nStand up man I will not win, but it's a really weird competition.
\n\nIs Blending the real point ? Don't know. Is there something to grab from Data ? Don't know.
\n\nWhish you the best, and I am sure you wille be among the best score, sure !
", "rawMarkdown": "konradb \n\nYou know what ? This competition is really oughta sight ! That's a weird competition : \n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/114893#latest-666521\n\nWhat a feeling having so much DownVotes...\n\nStand up man I will not win, but it's a really weird competition.\n\nIs Blending the real point ? Don't know. Is there something to grab from Data ? Don't know.\n\nWhish you the best, and I am sure you wille be among the best score, sure !", "votes": null }, { "id": "667974", "postDate": "11/07/2019 21:13:31", "content": "Hem, wot you mean ?
", "rawMarkdown": "Hem, wot you mean ?", "votes": null }, { "id": "667976", "postDate": "11/07/2019 21:15:41", "content": "This is Miles Davis, don't know what you are meaning.
\n\nJust Miles.
", "rawMarkdown": "This is Miles Davis, don't know what you are meaning.\n\nJust Miles.", "votes": null }, { "id": "667990", "postDate": "11/07/2019 21:24:57", "content": "It's been like a tsunami, but even if it was'nt for me it seems to shake up all around... And I can understand some frutration.
", "rawMarkdown": "It's been like a tsunami, but even if it was'nt for me it seems to shake up all around... And I can understand some frutration.", "votes": null }, { "id": "667995", "postDate": "11/07/2019 21:27:22", "content": "Good score !!!
", "rawMarkdown": "Good score !!!", "votes": null }, { "id": "668012", "postDate": "11/07/2019 21:50:51", "content": "I wish I could upvote you twice for that reference - one of my favourite BD tracks from his early era.
", "rawMarkdown": "I wish I could upvote you twice for that reference - one of my favourite BD tracks from his early era.", "votes": null }, { "id": "668015", "postDate": "11/07/2019 21:51:39", "content": "Thanks, that's very kind of you.
", "rawMarkdown": "Thanks, that's very kind of you.", "votes": null }, { "id": "668016", "postDate": "11/07/2019 21:52:33", "content": "The last mega drop - I busy myself with sth else for a day or two, go back to Kaggle, booom 100 spots down. Happened a few times before (with pub kernels usually), but never so fast :-)
", "rawMarkdown": "The last mega drop - I busy myself with sth else for a day or two, go back to Kaggle, booom 100 spots down. Happened a few times before (with pub kernels usually), but never so fast :-)", "votes": null }, { "id": "668873", "postDate": "11/09/2019 01:48:44", "content": "***My bittersweet experience of Blending:
\n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/116264#latest-668099
", "rawMarkdown": "***My bittersweet experience of Blending:\n***\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/116264#latest-668099", "votes": null }, { "id": "671517", "postDate": "11/12/2019 20:26:47", "content": "Yes, people should pay more attention to keeping tight correlation between their own CrossValidated score on training/holdout set, and their personal LB submission.
\n\nWhat training set split and CV methodology do you suggest?
\n\nDo you think the \"half and half\" (caret::createTimeSlices) timeseries rolling split causes leakage? esp. when one single day is split between training and holdout set?
Heewww ! It is so cold where you live.... Brrr Brrr Brr.
\n\nWait for tomorrow. Sun will be there, for sure !
", "votes": null, "replies": [] } ] }, { "id": 667533, "author_name": "manojprabhaakr", "author_url": "", "post_date": "11/07/2019 10:47:13", "content": "
Good atmosphere here !
\n\nThanks, I will be back ...
", "votes": null, "replies": [] } ] }, { "id": 667534, "author_name": "manojprabhaakr", "author_url": "", "post_date": "11/07/2019 10:47:54", "content": "Its the beginning of blending Season. It will be on till competition gets over.
", "votes": null, "replies": [] }, { "id": 667574, "author_name": "wispzero", "author_url": "", "post_date": "11/07/2019 12:00:07", "content": "Be more optimistic - that means different approaches could provide different results. And after you will add something unique to them (like features or processing) you will be able to increase your score at ease. Just don't spend any submission for blending right now - the blending guys did this for you.
", "votes": null, "replies": [ { "id": 667954, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 20:59:40", "content": "Hem ... Explain me what a blend can add to one competition, except but blending... Heeww (Homer Simpson saying).
", "votes": null, "replies": [] } ] }, { "id": 667621, "author_name": "hmendonca", "author_url": "", "post_date": "11/07/2019 13:11:32", "content": "Sorry mate! There is always a first one ;)\nBut it works as tutorial as well, for people that still don't know the power of the ensembles...
", "votes": null, "replies": [ { "id": 667912, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 20:10:25", "content": "But it works as tutorial as well
\n\nare you really sure ?
", "votes": null, "replies": [] }, { "id": 667990, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 21:24:57", "content": "It's been like a tsunami, but even if it was'nt for me it seems to shake up all around... And I can understand some frutration.
", "votes": null, "replies": [] } ] }, { "id": 667633, "author_name": "georsara1", "author_url": "", "post_date": "11/07/2019 13:30:35", "content": "\n\n\nBlending season is upon us
\n
...and it is actually making me really happy. When some get lured in the easy (but probably overfitted) path of LB climbing, others stick to their trusted models. It worked for me in IEEE Fraud Detection at least! :) \n
So what ? (Miles Davis)
\n\nhttps://www.youtube.com/watch?v=zqNTltOGh5c
\n\nlisten to that and keep cool 🤘
\n\nSure you will recover once.
", "votes": null, "replies": [] }, { "id": 667967, "author_name": "georsara1", "author_url": "", "post_date": "11/07/2019 21:10:12", "content": "That's *some *footage btw!
", "votes": null, "replies": [] }, { "id": 667976, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 21:15:41", "content": "This is Miles Davis, don't know what you are meaning.
\n\nJust Miles.
", "votes": null, "replies": [] }, { "id": 667995, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 21:27:22", "content": "Good score !!!
", "votes": null, "replies": [] }, { "id": 671517, "author_name": "smcinerney", "author_url": "", "post_date": "11/12/2019 20:26:47", "content": "Yes, people should pay more attention to keeping tight correlation between their own CrossValidated score on training/holdout set, and their personal LB submission.
\n\nWhat training set split and CV methodology do you suggest?
\n\nDo you think the \"half and half\" (caret::createTimeSlices) timeseries rolling split causes leakage? esp. when one single day is split between training and holdout set?
It's all over now... Baby Blue 🤜
\n\nhttps://www.youtube.com/watch?v=af7ngGxEusE
", "votes": null, "replies": [] }, { "id": 668012, "author_name": "konradb", "author_url": "", "post_date": "11/07/2019 21:50:51", "content": "I wish I could upvote you twice for that reference - one of my favourite BD tracks from his early era.
", "votes": null, "replies": [] } ] }, { "id": 667900, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 20:02:48", "content": "I've got a feeling it won't be the last :-/
\n\nwhat you mean ?
", "votes": null, "replies": [ { "id": 668016, "author_name": "konradb", "author_url": "", "post_date": "11/07/2019 21:52:33", "content": "The last mega drop - I busy myself with sth else for a day or two, go back to Kaggle, booom 100 spots down. Happened a few times before (with pub kernels usually), but never so fast :-)
", "votes": null, "replies": [] } ] }, { "id": 667907, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 20:08:53", "content": "", "votes": null, "replies": [] }, { "id": 667952, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 20:54:28", "content": "\n\nAllways been a follower and a supporter, really !
", "votes": null, "replies": [] }, { "id": 667958, "author_name": "jleecook", "author_url": "", "post_date": "11/07/2019 21:03:59", "content": "Hem, wot you mean ?
", "votes": null, "replies": [] } ] }, { "id": 667973, "author_name": "pfichou", "author_url": "", "post_date": "11/07/2019 21:12:46", "content": "\n\nYou know what ? This competition is really oughta sight ! That's a weird competition :
\n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/114893#latest-666521
\n\nWhat a feeling having so much DownVotes...
\n\nStand up man I will not win, but it's a really weird competition.
\n\nIs Blending the real point ? Don't know. Is there something to grab from Data ? Don't know.
\n\nWhish you the best, and I am sure you wille be among the best score, sure !
", "votes": null, "replies": [ { "id": 668015, "author_name": "konradb", "author_url": "", "post_date": "11/07/2019 21:51:39", "content": "Thanks, that's very kind of you.
", "votes": null, "replies": [] } ] }, { "id": 668873, "author_name": "grapestone5321", "author_url": "", "post_date": "11/09/2019 01:48:44", "content": "***My bittersweet experience of Blending:
\n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/116264#latest-668099
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "667486": "24h, one blending kernel => 90 spots collapse. I've got a feeling it won't be the last :-/", "667515": "", "667533": "", "667534": "Its the beginning of blending Season. It will be on till competition gets over.", "667574": "Be more optimistic - that means different approaches could provide different results. And after you will add something unique to them (like features or processing) you will be able to increase your score at ease. Just don't spend any submission for blending right now - the blending guys did this for you.", "667621": "Sorry mate! There is always a first one ;)\nBut it works as tutorial as well, for people that still don't know the power of the ensembles...", "667633": "> Blending season is upon us \n\n...and it is actually making me really happy. When some get lured in the easy (but probably overfitted) path of LB climbing, others stick to their trusted models. It worked for me in IEEE Fraud Detection at least! :) \n", "667882": "", "667900": "*I've got a feeling it won't be the last :-/*\n\nwhat you mean ?", "667907": "", "667912": "*But it works as tutorial as well*\n\nare you really sure ?", "667917": "So what ? (Miles Davis) \n\nhttps://www.youtube.com/watch?v=zqNTltOGh5c\n\nlisten to that and keep cool 🤘 \n\nSure you will recover once.", "667918": "Good atmosphere here ! \n\nThanks, I will be back ...", "667920": "Heewww ! It is so cold where you live.... Brrr Brrr Brr.\n\nWait for tomorrow. Sun will be there, for sure !", "667926": "It's all over now... Baby Blue 🤜 \n\nhttps://www.youtube.com/watch?v=af7ngGxEusE", "667952": "konradb \n\nAllways been a follower and a supporter, really !", "667954": "Hem ... Explain me what a blend can add to one competition, except but blending... Heeww (Homer Simpson saying).", "667958": "", "667967": "That's *some *footage btw!", "667973": "konradb \n\nYou know what ? This competition is really oughta sight ! That's a weird competition : \n\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/114893#latest-666521\n\nWhat a feeling having so much DownVotes...\n\nStand up man I will not win, but it's a really weird competition.\n\nIs Blending the real point ? Don't know. Is there something to grab from Data ? Don't know.\n\nWhish you the best, and I am sure you wille be among the best score, sure !", "667974": "Hem, wot you mean ?", "667976": "This is Miles Davis, don't know what you are meaning.\n\nJust Miles.", "667990": "It's been like a tsunami, but even if it was'nt for me it seems to shake up all around... And I can understand some frutration.", "667995": "Good score !!!", "668012": "I wish I could upvote you twice for that reference - one of my favourite BD tracks from his early era.", "668015": "Thanks, that's very kind of you.", "668016": "The last mega drop - I busy myself with sth else for a day or two, go back to Kaggle, booom 100 spots down. Happened a few times before (with pub kernels usually), but never so fast :-)", "668873": "***My bittersweet experience of Blending:\n***\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/116264#latest-668099", "671517": "Yes, people should pay more attention to keeping tight correlation between their own CrossValidated score on training/holdout set, and their personal LB submission.\n\nWhat training set split and CV methodology do you suggest?\n\nDo you think the \"half and half\" (`caret::createTimeSlices`) timeseries rolling split causes leakage? esp. when one single day is split between training and holdout set?" }, "source": "meta" }![\"\"](\"https://media3.webcollage.net/1f00857c975f3b7fb488793c018c23adb73346c1?response-content-type=image%2Fpng&AWSAccessKeyId=AKIAIIE5CHZ4PRWSLYKQ&Expires=1893466398&Signature=W6NKXF7sO8%2B0WIutemYirVo1dQE%3D\")