{ "id": 230404, "title": "Aiming for a sub-3 score πŸ˜„", "url": "/competitions/indoor-location-navigation/discussion/230404", "author_name": "", "post_date": "2021-04-03T13:50:33.572996600Z", "votes": 18, "comment_count": 10, "views": 0, "content": "

@mamasinkgs who is well on this way to a sub-3 score had this suggestion for us to get better scores:
\n(1) EDA by yourself.
\n(2) Understand good kernels (regardless of their score).
\n(Source: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796)
\n\\n
\nWhat do you think are (a) some points in EDA that we should focus on/may have been missed and
\n(b) some understated notebooks publishes that we should be looking at?
\nLets discuss and share what we believe are worth more attention and hoping to see many of us in sub-3 score range soon!!
\nSome points to get this discussion thread started:
\n\\n
\n1) Leak and how to take advantage of it! very grateful for @tomooinubushi for sharing with all!!
\nDiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228898
\nCode: https://www.kaggle.com/tomooinubushi/postprocessing-based-on-leakage
\nEven @mamasinkgs acknowledged it helped and there is much more we can do with this: https://www.kaggle.com/c/indoor-location-navigation/discussion/230096
\n\\n
\n2) @kokitanisaka notebook on how to create: one dataset for one model
\ndiscussion https://www.kaggle.com/c/indoor-location-navigation/discussion/222580
\nCode: https://www.kaggle.com/kokitanisaka/create-unified-wifi-features-example

\n

\\n
\n3) @nooblife CV strategy is very useful
\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228199
\n\\n
\n4) I agree with @nigelhenry that \"… when the submissions are the other way around (rely on accelerometer and gyroscope corrected by wifi) then I suspect it should be accurate to 1.5 - 2 strides which I believe is between 1 and 2 by the error metric.\"
\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796
\n\\n
\nEDA: @andradaolteanu 's notebooks are always a great starting point!
\nhttps://www.kaggle.com/andradaolteanu/indoor-navigation-complete-data-understanding
\nHowever, I may be looking at the more obvious ones… any hidden gems you recommend?
\nSee you all at sub-3 scores soon πŸ˜„πŸ‘πŸ™Œ

", "messages": [ { "id": "1261833", "postDate": "04/03/2021 13:50:33", "content": "

@mamasinkgs who is well on this way to a sub-3 score had this suggestion for us to get better scores:
\n(1) EDA by yourself.
\n(2) Understand good kernels (regardless of their score).
\n(Source: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796)
\n\\n
\nWhat do you think are (a) some points in EDA that we should focus on/may have been missed and
\n(b) some understated notebooks publishes that we should be looking at?
\nLets discuss and share what we believe are worth more attention and hoping to see many of us in sub-3 score range soon!!
\nSome points to get this discussion thread started:
\n\\n
\n1) Leak and how to take advantage of it! very grateful for @tomooinubushi for sharing with all!!
\nDiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228898
\nCode: https://www.kaggle.com/tomooinubushi/postprocessing-based-on-leakage
\nEven @mamasinkgs acknowledged it helped and there is much more we can do with this: https://www.kaggle.com/c/indoor-location-navigation/discussion/230096
\n\\n
\n2) @kokitanisaka notebook on how to create: one dataset for one model
\ndiscussion https://www.kaggle.com/c/indoor-location-navigation/discussion/222580
\nCode: https://www.kaggle.com/kokitanisaka/create-unified-wifi-features-example

\n

\\n
\n3) @nooblife CV strategy is very useful
\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228199
\n\\n
\n4) I agree with @nigelhenry that \"… when the submissions are the other way around (rely on accelerometer and gyroscope corrected by wifi) then I suspect it should be accurate to 1.5 - 2 strides which I believe is between 1 and 2 by the error metric.\"
\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796
\n\\n
\nEDA: @andradaolteanu 's notebooks are always a great starting point!
\nhttps://www.kaggle.com/andradaolteanu/indoor-navigation-complete-data-understanding
\nHowever, I may be looking at the more obvious ones… any hidden gems you recommend?
\nSee you all at sub-3 scores soon πŸ˜„πŸ‘πŸ™Œ

", "rawMarkdown": "mamasinkgs who is well on this way to a sub-3 score had this suggestion for us to get better scores:\n\n(1) EDA by yourself.\n(2) Understand good kernels (regardless of their score).\n(Source: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796)\n\n\\n\n\nWhat do you think are (a) some points in EDA that we should focus on/may have been missed and \n(b) some understated notebooks publishes that we should be looking at?\n\nLets discuss and share what we believe are worth more attention and hoping to see many of us in sub-3 score range soon!!\n\n\nSome points to get this discussion thread started:\n\n\\n\n\n1) Leak and how to take advantage of it! very grateful for @tomooinubushi for sharing with all!!\nDiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228898\nCode: https://www.kaggle.com/tomooinubushi/postprocessing-based-on-leakage\n\nEven @mamasinkgs acknowledged it helped and there is much more we can do with this: https://www.kaggle.com/c/indoor-location-navigation/discussion/230096\n\\n\n2) @kokitanisaka notebook on how to create: one dataset for one model\ndiscussion https://www.kaggle.com/c/indoor-location-navigation/discussion/222580\nCode: https://www.kaggle.com/kokitanisaka/create-unified-wifi-features-example\n \n\\n\n3) @nooblife CV strategy is very useful \ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228199\n\n\\n\n4) I agree with @nigelhenry that \"... when the submissions are the other way around (rely on accelerometer and gyroscope corrected by wifi) then I suspect it should be accurate to 1.5 - 2 strides which I believe is between 1 and 2 by the error metric.\"\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796\n\n\\n\nEDA: @andradaolteanu 's notebooks are always a great starting point!\nhttps://www.kaggle.com/andradaolteanu/indoor-navigation-complete-data-understanding\n\nHowever, I may be looking at the more obvious ones... any hidden gems you recommend?\n\nSee you all at sub-3 scores soon πŸ˜„πŸ‘πŸ™Œ", "votes": null }, { "id": "1262761", "postDate": "04/04/2021 17:27:02", "content": "

@kmldas good accounting of the current state of this competition. I've been wrestling a bit with the various pre-processing and areas that have been pointed out.. such as fixed timestamps and incorrect data fields and working those into my process--or not.

", "rawMarkdown": "kmldas good accounting of the current state of this competition. I've been wrestling a bit with the various pre-processing and areas that have been pointed out.. such as fixed timestamps and incorrect data fields and working those into my process--or not.", "votes": null }, { "id": "1266991", "postDate": "04/08/2021 08:20:02", "content": "

Some good old fashioned smoke signals observed last night as two leading contenders communicated via LB team name changes.

", "rawMarkdown": "Some good old fashioned smoke signals observed last night as two leading contenders communicated via LB team name changes.", "votes": null }, { "id": "1267258", "postDate": "04/08/2021 12:20:26", "content": "

sorry @jbomitchell … did not understand what you meant

", "rawMarkdown": "sorry @jbomitchell ... did not understand what you meant", "votes": null }, { "id": "1267267", "postDate": "04/08/2021 12:25:21", "content": "

@kmldas - the name of the top team on the leaderboard was changed to \"Youri Matiounine please check e-mail\". These two teams have now merged, bringing @mamasinkgs @ymatioun together into one superteam.

", "rawMarkdown": "kmldas - the name of the top team on the leaderboard was changed to \"Youri Matiounine please check e-mail\". These two teams have now merged, bringing @mamasinkgs @ymatioun together into one superteam.", "votes": null }, { "id": "1267304", "postDate": "04/08/2021 12:58:41", "content": "

Doh! Now it's getting even more competitive. I think sub 3 is the goal and probably achievable. Might even be necessary to get in winners circle at the end.

", "rawMarkdown": "Doh! Now it's getting even more competitive. I think sub 3 is the goal and probably achievable. Might even be necessary to get in winners circle at the end.", "votes": null }, { "id": "1267332", "postDate": "04/08/2021 13:22:47", "content": "

This new team was unexpected :) I'm quite sure that just averaging their current prediction, will give them a sub3 score!

", "rawMarkdown": "This new team was unexpected :) I'm quite sure that just averaging their current prediction, will give them a sub3 score!", "votes": null }, { "id": "1267427", "postDate": "04/08/2021 14:05:27", "content": "

I had missed the name change… very observant @jbomitchell πŸ˜„
\nWith so many individual competitors here, expect many more mergers in the next few weeks

\n

lets see how low we can go!! I think the gold medal teams may be in the 2.25 to 2.5 band
\n@zidmie had said \"My bet is for a gold zone of 2.6-2.9 (public LB scores).\"
\nsource: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796

\n

what are your (updated) predictions…

", "rawMarkdown": "I had missed the name change... very observant @jbomitchell πŸ˜„\nWith so many individual competitors here, expect many more mergers in the next few weeks\n\nlets see how low we can go!! I think the gold medal teams may be in the 2.25 to 2.5 band\n@zidmie had said \"My bet is for a gold zone of 2.6-2.9 (public LB scores).\"\nsource: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796\n\nwhat are your (updated) predictions...", "votes": null }, { "id": "1267618", "postDate": "04/08/2021 16:36:59", "content": "

My friend said a top score on the date 1 month before the deadline (4/17) is a good estimation of the score that can get a gold medal.
\nBTW, we got 2.8 by teaming up 😍😍😍, sub 3 achieved :)

", "rawMarkdown": "My friend said a top score on the date 1 month before the deadline (4/17) is a good estimation of the score that can get a gold medal.\nBTW, we got 2.8 by teaming up 😍😍😍, sub 3 achieved :)", "votes": null }, { "id": "1276861", "postDate": "04/18/2021 05:08:43", "content": "

Gold spot for below 2.483!! a month before the deadline -- you are making it tough for your fellow competitors πŸ˜„πŸ˜„!!

\n

CONGRATULATIONS @mamasinkgs & @ymatioun; you are now below the 2.5 barrier
\nA Public LB score of 2.483 is just unbelievable !!

\n

With @seuguh possibly merging based on your team name, wishing to see you below 2 soon!! I am looking towards you to see how far we can go in this competition!!

", "rawMarkdown": "Gold spot for below 2.483!! a month before the deadline -- you are making it tough for your fellow competitors πŸ˜„πŸ˜„!!\n\nCONGRATULATIONS @mamasinkgs & @ymatioun; you are now below the 2.5 barrier \nA Public LB score of 2.483 is just unbelievable !!\n\nWith @seuguh possibly merging based on your team name, wishing to see you below 2 soon!! I am looking towards you to see how far we can go in this competition!!", "votes": null }, { "id": "1276875", "postDate": "04/18/2021 05:36:45", "content": "

Thanks, we'll keep updating our strategies!

", "rawMarkdown": "Thanks, we'll keep updating our strategies!", "votes": null } ], "comments": [ { "id": 1262761, "author_name": "crained", "author_url": "", "post_date": "04/04/2021 17:27:02", "content": "

@kmldas good accounting of the current state of this competition. I've been wrestling a bit with the various pre-processing and areas that have been pointed out.. such as fixed timestamps and incorrect data fields and working those into my process--or not.

", "votes": null, "replies": [] }, { "id": 1266991, "author_name": "jbomitchell", "author_url": "", "post_date": "04/08/2021 08:20:02", "content": "

Some good old fashioned smoke signals observed last night as two leading contenders communicated via LB team name changes.

", "votes": null, "replies": [ { "id": 1267258, "author_name": "kmldas", "author_url": "", "post_date": "04/08/2021 12:20:26", "content": "

sorry @jbomitchell … did not understand what you meant

", "votes": null, "replies": [] }, { "id": 1267267, "author_name": "jbomitchell", "author_url": "", "post_date": "04/08/2021 12:25:21", "content": "

@kmldas - the name of the top team on the leaderboard was changed to \"Youri Matiounine please check e-mail\". These two teams have now merged, bringing @mamasinkgs @ymatioun together into one superteam.

", "votes": null, "replies": [] }, { "id": 1267304, "author_name": "crained", "author_url": "", "post_date": "04/08/2021 12:58:41", "content": "

Doh! Now it's getting even more competitive. I think sub 3 is the goal and probably achievable. Might even be necessary to get in winners circle at the end.

", "votes": null, "replies": [] }, { "id": 1267332, "author_name": "zidmie", "author_url": "", "post_date": "04/08/2021 13:22:47", "content": "

This new team was unexpected :) I'm quite sure that just averaging their current prediction, will give them a sub3 score!

", "votes": null, "replies": [] }, { "id": 1267427, "author_name": "kmldas", "author_url": "", "post_date": "04/08/2021 14:05:27", "content": "

I had missed the name change… very observant @jbomitchell πŸ˜„
\nWith so many individual competitors here, expect many more mergers in the next few weeks

\n

lets see how low we can go!! I think the gold medal teams may be in the 2.25 to 2.5 band
\n@zidmie had said \"My bet is for a gold zone of 2.6-2.9 (public LB scores).\"
\nsource: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796

\n

what are your (updated) predictions…

", "votes": null, "replies": [] }, { "id": 1267618, "author_name": "mamasinkgs", "author_url": "", "post_date": "04/08/2021 16:36:59", "content": "

My friend said a top score on the date 1 month before the deadline (4/17) is a good estimation of the score that can get a gold medal.
\nBTW, we got 2.8 by teaming up 😍😍😍, sub 3 achieved :)

", "votes": null, "replies": [] }, { "id": 1276861, "author_name": "kmldas", "author_url": "", "post_date": "04/18/2021 05:08:43", "content": "

Gold spot for below 2.483!! a month before the deadline -- you are making it tough for your fellow competitors πŸ˜„πŸ˜„!!

\n

CONGRATULATIONS @mamasinkgs & @ymatioun; you are now below the 2.5 barrier
\nA Public LB score of 2.483 is just unbelievable !!

\n

With @seuguh possibly merging based on your team name, wishing to see you below 2 soon!! I am looking towards you to see how far we can go in this competition!!

", "votes": null, "replies": [] }, { "id": 1276875, "author_name": "mamasinkgs", "author_url": "", "post_date": "04/18/2021 05:36:45", "content": "

Thanks, we'll keep updating our strategies!

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1261833": "mamasinkgs who is well on this way to a sub-3 score had this suggestion for us to get better scores:\n\n(1) EDA by yourself.\n(2) Understand good kernels (regardless of their score).\n(Source: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796)\n\n\\n\n\nWhat do you think are (a) some points in EDA that we should focus on/may have been missed and \n(b) some understated notebooks publishes that we should be looking at?\n\nLets discuss and share what we believe are worth more attention and hoping to see many of us in sub-3 score range soon!!\n\n\nSome points to get this discussion thread started:\n\n\\n\n\n1) Leak and how to take advantage of it! very grateful for @tomooinubushi for sharing with all!!\nDiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228898\nCode: https://www.kaggle.com/tomooinubushi/postprocessing-based-on-leakage\n\nEven @mamasinkgs acknowledged it helped and there is much more we can do with this: https://www.kaggle.com/c/indoor-location-navigation/discussion/230096\n\\n\n2) @kokitanisaka notebook on how to create: one dataset for one model\ndiscussion https://www.kaggle.com/c/indoor-location-navigation/discussion/222580\nCode: https://www.kaggle.com/kokitanisaka/create-unified-wifi-features-example\n \n\\n\n3) @nooblife CV strategy is very useful \ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228199\n\n\\n\n4) I agree with @nigelhenry that \"... when the submissions are the other way around (rely on accelerometer and gyroscope corrected by wifi) then I suspect it should be accurate to 1.5 - 2 strides which I believe is between 1 and 2 by the error metric.\"\ndiscussion: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796\n\n\\n\nEDA: @andradaolteanu 's notebooks are always a great starting point!\nhttps://www.kaggle.com/andradaolteanu/indoor-navigation-complete-data-understanding\n\nHowever, I may be looking at the more obvious ones... any hidden gems you recommend?\n\nSee you all at sub-3 scores soon πŸ˜„πŸ‘πŸ™Œ", "1262761": "kmldas good accounting of the current state of this competition. I've been wrestling a bit with the various pre-processing and areas that have been pointed out.. such as fixed timestamps and incorrect data fields and working those into my process--or not.", "1266991": "Some good old fashioned smoke signals observed last night as two leading contenders communicated via LB team name changes.", "1267258": "sorry @jbomitchell ... did not understand what you meant", "1267267": "kmldas - the name of the top team on the leaderboard was changed to \"Youri Matiounine please check e-mail\". These two teams have now merged, bringing @mamasinkgs @ymatioun together into one superteam.", "1267304": "Doh! Now it's getting even more competitive. I think sub 3 is the goal and probably achievable. Might even be necessary to get in winners circle at the end.", "1267332": "This new team was unexpected :) I'm quite sure that just averaging their current prediction, will give them a sub3 score!", "1267427": "I had missed the name change... very observant @jbomitchell πŸ˜„\nWith so many individual competitors here, expect many more mergers in the next few weeks\n\nlets see how low we can go!! I think the gold medal teams may be in the 2.25 to 2.5 band\n@zidmie had said \"My bet is for a gold zone of 2.6-2.9 (public LB scores).\"\nsource: https://www.kaggle.com/c/indoor-location-navigation/discussion/228796\n\nwhat are your (updated) predictions...", "1267618": "My friend said a top score on the date 1 month before the deadline (4/17) is a good estimation of the score that can get a gold medal.\nBTW, we got 2.8 by teaming up 😍😍😍, sub 3 achieved :)", "1276861": "Gold spot for below 2.483!! a month before the deadline -- you are making it tough for your fellow competitors πŸ˜„πŸ˜„!!\n\nCONGRATULATIONS @mamasinkgs & @ymatioun; you are now below the 2.5 barrier \nA Public LB score of 2.483 is just unbelievable !!\n\nWith @seuguh possibly merging based on your team name, wishing to see you below 2 soon!! I am looking towards you to see how far we can go in this competition!!", "1276875": "Thanks, we'll keep updating our strategies!" }, "source": "meta" }