{ "id": 409509, "title": "Unit unification and resampling", "url": "/competitions/tlvmc-parkinsons-freezing-gait-prediction/discussion/409509", "author_name": "Lingduo Wang", "post_date": "2023-05-11T11:17:07.851000", "votes": 0, "comment_count": 9, "views": 0, "content": "

Hello!

\n

According to the data description, the unit and sampling rate of defog and tdcsfog are inconsistent. I would like to ask if you have used unit unification and resampling?

\n

I think these two modules are very correct and should be able to produce positive results. But when I was experimenting, the results of these two modules were not optimistic.

\n

I would like to ask how you deal with units and sampling rates.

\n

Thank you for your attention! ;)

", "messages": [ { "id": 2256867, "postDate": "2023-05-12T19:43:27.390Z", "content": "

try using the data as is - most models will be just fine with it.

", "rawMarkdown": "try using the data as is - most models will be just fine with it.", "votes": 1, "replies": [ { "id": 2257326, "postDate": "2023-05-13T08:56:16.780Z", "content": "

isn't it, in principle, equivalent of having two models?

", "rawMarkdown": "isn't it, in principle, equivalent of having two models?", "replies": [ { "id": 2257708, "postDate": "2023-05-13T15:51:58.300Z", "content": "

Well, it depends on your feature engineering.
\nBut it doesn't have to be.

", "rawMarkdown": "Well, it depends on your feature engineering.\nBut it doesn't have to be." }, { "id": 2259829, "postDate": "2023-05-15T09:03:14.250Z", "content": "

well, are you suggesting not \"using data as is\"? : )

", "rawMarkdown": "well, are you suggesting not \"using data as is\"? : )" } ] } ] }, { "id": 2256012, "postDate": "2023-05-12T07:40:39.907Z", "content": "

my experience is also a worse score when unifying datasets and training a single model

", "rawMarkdown": "my experience is also a worse score when unifying datasets and training a single model", "votes": 1 }, { "id": 2256274, "postDate": "2023-05-12T11:15:00.600Z", "content": "

It is surprising that two separate models do better than training with a single model for the tDCS and DeFOG datasets. My assumption is that a single model that works well on all 3 datasets will likely to be robust and perform well, avoiding overfitting.

", "rawMarkdown": "It is surprising that two separate models do better than training with a single model for the tDCS and DeFOG datasets. My assumption is that a single model that works well on all 3 datasets will likely to be robust and perform well, avoiding overfitting.", "votes": 2, "replies": [ { "id": 2257054, "postDate": "2023-05-13T01:06:52.207Z", "content": "

Thank you so much for your reply! I will try to train them separately with unit unification and resampling！:)

", "rawMarkdown": "Thank you so much for your reply! I will try to train them separately with unit unification and resampling！:)" }, { "id": 2259053, "postDate": "2023-05-14T16:50:10.837Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 2259171, "postDate": "2023-05-14T18:23:57.817Z", "content": "

The tilt correction method referred to in this paper (reference 16) might help to deal with some of these issues:
\n\"Therefore, a commonly used tilt correction procedure proposed by Moe-Nilssen et al. [16] was applied to all directions before the extraction of accelerometry-based features.\"
\nhttps://www.mdpi.com/1424-8220/21/21/7385 Gerard Aristizábal Pla et al. (Sensors, 2021)

", "rawMarkdown": "The tilt correction method referred to in this paper (reference 16) might help to deal with some of these issues:\n\"Therefore, a commonly used tilt correction procedure proposed by Moe-Nilssen et al. [16] was applied to all directions before the extraction of accelerometry-based features.\" \nhttps://www.mdpi.com/1424-8220/21/21/7385 Gerard Aristizábal Pla et al. (Sensors, 2021)", "votes": 4 } ] } ] }, { "id": 2254950, "postDate": "2023-05-11T11:17:07.853Z", "content": "

Hello!

\n

According to the data description, the unit and sampling rate of defog and tdcsfog are inconsistent. I would like to ask if you have used unit unification and resampling?

\n

I think these two modules are very correct and should be able to produce positive results. But when I was experimenting, the results of these two modules were not optimistic.

\n

I would like to ask how you deal with units and sampling rates.

\n

Thank you for your attention! ;)

", "rawMarkdown": "Hello! \n\nAccording to the data description, the unit and sampling rate of defog and tdcsfog are inconsistent. I would like to ask if you have used unit unification and resampling?\n\n I think these two modules are very correct and should be able to produce positive results. But when I was experimenting, the results of these two modules were not optimistic. \n\nI would like to ask how you deal with units and sampling rates. \n\nThank you for your attention! ;)" } ], "comments": [ { "id": 2256867, "author_name": "Aviv Levi", "author_url": "", "post_date": "2023-05-12T19:43:27.390000", "content": "

try using the data as is - most models will be just fine with it.

", "votes": 1, "replies": [ { "id": 2257326, "author_name": "Alberto Annoni", "author_url": "", "post_date": "2023-05-13T08:56:16.780000", "content": "

isn't it, in principle, equivalent of having two models?

", "votes": 0, "replies": [ { "id": 2257708, "author_name": "Aviv Levi", "author_url": "", "post_date": "2023-05-13T15:51:58.300000", "content": "

Well, it depends on your feature engineering.
\nBut it doesn't have to be.

", "votes": 0, "replies": [] }, { "id": 2259829, "author_name": "Alberto Annoni", "author_url": "", "post_date": "2023-05-15T09:03:14.250000", "content": "

well, are you suggesting not \"using data as is\"? : )

", "votes": 0, "replies": [] } ] } ] }, { "id": 2256012, "author_name": "Alberto Annoni", "author_url": "", "post_date": "2023-05-12T07:40:39.907000", "content": "

my experience is also a worse score when unifying datasets and training a single model

", "votes": 1, "replies": [] }, { "id": 2256274, "author_name": "Jeff Hausdorff", "author_url": "", "post_date": "2023-05-12T11:15:00.600000", "content": "

It is surprising that two separate models do better than training with a single model for the tDCS and DeFOG datasets. My assumption is that a single model that works well on all 3 datasets will likely to be robust and perform well, avoiding overfitting.

", "votes": 2, "replies": [ { "id": 2257054, "author_name": "Lingduo Wang", "author_url": "", "post_date": "2023-05-13T01:06:52.207000", "content": "

Thank you so much for your reply! I will try to train them separately with unit unification and resampling！:)

", "votes": 0, "replies": [] }, { "id": 2259053, "author_name": "", "author_url": "", "post_date": "2023-05-14T16:50:10.837000", "content": "", "votes": 0, "replies": [ { "id": 2259171, "author_name": "Jeff Hausdorff", "author_url": "", "post_date": "2023-05-14T18:23:57.817000", "content": "

The tilt correction method referred to in this paper (reference 16) might help to deal with some of these issues:
\n\"Therefore, a commonly used tilt correction procedure proposed by Moe-Nilssen et al. [16] was applied to all directions before the extraction of accelerometry-based features.\"
\nhttps://www.mdpi.com/1424-8220/21/21/7385 Gerard Aristizábal Pla et al. (Sensors, 2021)

", "votes": 4, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2256867": "try using the data as is - most models will be just fine with it.", "2256012": "my experience is also a worse score when unifying datasets and training a single model", "2256274": "It is surprising that two separate models do better than training with a single model for the tDCS and DeFOG datasets. My assumption is that a single model that works well on all 3 datasets will likely to be robust and perform well, avoiding overfitting.", "2254950": "Hello! \n\nAccording to the data description, the unit and sampling rate of defog and tdcsfog are inconsistent. I would like to ask if you have used unit unification and resampling?\n\n I think these two modules are very correct and should be able to produce positive results. But when I was experimenting, the results of these two modules were not optimistic. \n\nI would like to ask how you deal with units and sampling rates. \n\nThank you for your attention! ;)" } }