\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F317344%2F09f038c2b5806f82b2b699837d14b6fe%2Flb%2012-06.png?generation=1701834710456805&alt=media\")
Top teams are leaving us behind.
\n
old one(11/30) is here
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F317344%2F99d9162fa58560215a53b3907e3a3507%2Flb.png?generation=1701329140575892&alt=media\")
Top teams are leaving us behind.
\n
old one(11/30) is here
\n
I think this is the border where the model stops approximating the mean and starts to really understand what RNA is, can't wait to see what hosts think after competition deadline :]
", "rawMarkdown": "I think this is the border where the model stops approximating the mean and starts to really understand what RNA is, can't wait to see what hosts think after competition deadline :]", "votes": 3, "replies": [ { "id": 2550896, "postDate": "2023-12-06T11:34:28.140Z", "content": "by border do you mean one specific gap? or all 3?
\nexcited to see what ya'll came up with as well!
I'm talking about 0.142 border, but it's all speculations really :D
", "rawMarkdown": "I'm talking about 0.142 border, but it's all speculations really :D ", "votes": 1 } ] } ] }, { "id": 2550597, "postDate": "2023-12-06T06:12:39.500Z", "content": "Yes! To my knowledge, all this bariers have explanation, but we should wait until the competition ends)
", "rawMarkdown": "Yes! To my knowledge, all this bariers have explanation, but we should wait until the competition ends)", "votes": 3, "replies": [ { "id": 2550895, "postDate": "2023-12-06T11:34:02.597Z", "content": "I think your team has a lot of domain knowledge, and am looking forward to hearing the solution soon
", "rawMarkdown": "I think your team has a lot of domain knowledge, and am looking forward to hearing the solution soon", "votes": 2, "replies": [ { "id": 2550991, "postDate": "2023-12-06T12:54:44.360Z", "content": "unfortunately, our domain knowledge is mainly about DNA, not RNA( Tasks are very different(
", "rawMarkdown": "unfortunately, our domain knowledge is mainly about DNA, not RNA( Tasks are very different( ", "votes": 2 } ] } ] }, { "id": 2550981, "postDate": "2023-12-06T12:46:04.733Z", "content": "By the way, I watched some lecture from Rhiju Das and if I remember correctly, they said they have a specific number in mind \"what they think, what result is enough for the models to be useful\", would love to hear feedback on that too
", "rawMarkdown": "By the way, I watched some lecture from Rhiju Das and if I remember correctly, they said they have a specific number in mind \"what they think, what result is enough for the models to be useful\", would love to hear feedback on that too", "votes": 1, "replies": [ { "id": 2551294, "postDate": "2023-12-06T15:52:34.847Z", "rawMarkdown": "", "votes": 1, "isDeleted": true, "replies": [ { "id": 2551373, "postDate": "2023-12-06T16:56:57.270Z", "content": "I don't think the structures predicted in the post are meaningful and helpfull.
\nFor example this one is a total mess
\n ![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F732016%2Fd2330f66dc9bc495027de66ae5aa84d3%2Fphoto_2023-11-04_23-43-18.jpg?generation=1701881766497911&alt=media\")
Even more Fun Fact - you have Dieter in your team! So never fear being left behind - it ain't over till the fat private LB sings.
\nAlso hardware and ensembles might be helping at the top?
Actually, my teammate @kfujikawa 's model is the our best single model so far.
\nBut this is a team, and this result would not have been possible without anyone missing.
but you guys always end up at 2nd place at the end -)
\nthe private might be different then public …
Have some ideas on the causes of these gaps, but not sure if we have enough time to train for them at this point
", "rawMarkdown": "Have some ideas on the causes of these gaps, but not sure if we have enough time to train for them at this point", "votes": 2 }, { "id": 2550939, "postDate": "2023-12-06T12:16:44.537Z", "content": "Possibly could be meaningful. But if we take 15 random numbers between 0.13500 and 0.14200, maybe we should expect to see some similarly uneven gaps just by the randomness of statistics and probability? So I'd be wary of over-interpreting a small sample of numbers with a complicated narrative.
", "rawMarkdown": "Possibly could be meaningful. But if we take 15 random numbers between 0.13500 and 0.14200, maybe we should expect to see some similarly uneven gaps just by the randomness of statistics and probability? So I'd be wary of over-interpreting a small sample of numbers with a complicated narrative." } ], "comments": [ { "id": 2550858, "author_name": "slime", "author_url": "", "post_date": "2023-12-06T10:36:43.290000", "content": "I think this is the border where the model stops approximating the mean and starts to really understand what RNA is, can't wait to see what hosts think after competition deadline :]
", "votes": 3, "replies": [ { "id": 2550896, "author_name": "sroger", "author_url": "", "post_date": "2023-12-06T11:34:28.140000", "content": "by border do you mean one specific gap? or all 3?
\nexcited to see what ya'll came up with as well!
I'm talking about 0.142 border, but it's all speculations really :D
", "votes": 1, "replies": [] } ] } ] }, { "id": 2550597, "author_name": "Penzar Dmitry", "author_url": "", "post_date": "2023-12-06T06:12:39.500000", "content": "Yes! To my knowledge, all this bariers have explanation, but we should wait until the competition ends)
", "votes": 3, "replies": [ { "id": 2550895, "author_name": "ONODERA", "author_url": "", "post_date": "2023-12-06T11:34:02.597000", "content": "I think your team has a lot of domain knowledge, and am looking forward to hearing the solution soon
", "votes": 2, "replies": [ { "id": 2550991, "author_name": "Penzar Dmitry", "author_url": "", "post_date": "2023-12-06T12:54:44.360000", "content": "unfortunately, our domain knowledge is mainly about DNA, not RNA( Tasks are very different(
", "votes": 2, "replies": [] } ] } ] }, { "id": 2550981, "author_name": "slime", "author_url": "", "post_date": "2023-12-06T12:46:04.733000", "content": "By the way, I watched some lecture from Rhiju Das and if I remember correctly, they said they have a specific number in mind \"what they think, what result is enough for the models to be useful\", would love to hear feedback on that too
", "votes": 1, "replies": [ { "id": 2551294, "author_name": "", "author_url": "", "post_date": "2023-12-06T15:52:34.847000", "content": "", "votes": 1, "replies": [ { "id": 2551373, "author_name": "Penzar Dmitry", "author_url": "", "post_date": "2023-12-06T16:56:57.270000", "content": "I don't think the structures predicted in the post are meaningful and helpfull.
\nFor example this one is a total mess
\n
Even more Fun Fact - you have Dieter in your team! So never fear being left behind - it ain't over till the fat private LB sings.
\nAlso hardware and ensembles might be helping at the top?
Actually, my teammate @kfujikawa 's model is the our best single model so far.
\nBut this is a team, and this result would not have been possible without anyone missing.
but you guys always end up at 2nd place at the end -)
\nthe private might be different then public …
Have some ideas on the causes of these gaps, but not sure if we have enough time to train for them at this point
", "votes": 2, "replies": [] }, { "id": 2550939, "author_name": "John Mitchell", "author_url": "", "post_date": "2023-12-06T12:16:44.537000", "content": "Possibly could be meaningful. But if we take 15 random numbers between 0.13500 and 0.14200, maybe we should expect to see some similarly uneven gaps just by the randomness of statistics and probability? So I'd be wary of over-interpreting a small sample of numbers with a complicated narrative.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2550482": "Top teams are leaving us behind.\n\n\n\nold one(11/30) is here\n", "2550858": "I think this is the border where the model stops approximating the mean and starts to really understand what RNA is, can't wait to see what hosts think after competition deadline :]", "2550597": "Yes! To my knowledge, all this bariers have explanation, but we should wait until the competition ends)", "2550981": "By the way, I watched some lecture from Rhiju Das and if I remember correctly, they said they have a specific number in mind \"what they think, what result is enough for the models to be useful\", would love to hear feedback on that too", "2550675": "Even more Fun Fact - you have Dieter in your team! So never fear being left behind - it ain't over till the fat private LB sings. \nAlso hardware and ensembles might be helping at the top?", "2550807": "Have some ideas on the causes of these gaps, but not sure if we have enough time to train for them at this point", "2550939": "Possibly could be meaningful. But if we take 15 random numbers between 0.13500 and 0.14200, maybe we should expect to see some similarly uneven gaps just by the randomness of statistics and probability? So I'd be wary of over-interpreting a small sample of numbers with a complicated narrative." } }