Could you guys share and discuss some new ideas? (Without compromising the competition's and Kaggle's rules of course). I have been stuck for a couple of weeks now and none of my new ideas have been able to improve the performance - how can one use features like the signal to noise, etc?
", "messages": [ { "id": "2492926", "postDate": "10/23/2023 03:22:38", "content": "Could you guys share and discuss some new ideas? (Without compromising the competition's and Kaggle's rules of course). I have been stuck for a couple of weeks now and none of my new ideas have been able to improve the performance - how can one use features like the signal to noise, etc?
", "rawMarkdown": "Could you guys share and discuss some new ideas? (Without compromising the competition's and Kaggle's rules of course). I have been stuck for a couple of weeks now and none of my new ideas have been able to improve the performance - how can one use features like the signal to noise, etc?", "votes": null }, { "id": "2492980", "postDate": "10/23/2023 04:53:53", "content": "No free lunch theorem! Keep grinding :)
", "rawMarkdown": "No free lunch theorem! Keep grinding :)", "votes": null }, { "id": "2493074", "postDate": "10/23/2023 06:28:58", "content": "I tried to use conv but my training stuck at some point not improve i dont know why have you face this i wonder that if conv layers are effective or not here?
", "rawMarkdown": "I tried to use conv but my training stuck at some point not improve i dont know why have you face this i wonder that if conv layers are effective or not here?", "votes": null }, { "id": "2493733", "postDate": "10/23/2023 14:41:50", "content": "Yes I faced the same issue
", "rawMarkdown": "Yes I faced the same issue", "votes": null }, { "id": "2494015", "postDate": "10/23/2023 17:08:37", "content": "You could try pulling in more training data from RMDB or EternaBench or any other sources of RNA structure data you can find.
\nA really good model needs to understand 3D structure, which is a tall challenge. Has anyone explored AlphaFold style Evoformer modules?
", "rawMarkdown": "You could try pulling in more training data from [RMDB](https://rmdb.stanford.edu/) or [EternaBench](https://github.com/eternagame/EternaBench) or any other sources of RNA structure data you can find.\n\nA really good model needs to understand 3D structure, which is a tall challenge. Has anyone explored AlphaFold style Evoformer modules?", "votes": null }, { "id": "2494024", "postDate": "10/23/2023 17:16:15", "content": "How to access the data? I signed up but could not get access to it. No,How to access the data? I signed up but could not get access to it. No I have not tried AF style models
", "rawMarkdown": "How to access the data? I signed up but could not get access to it. No,How to access the data? I signed up but could not get access to it. No I have not tried AF style models", "votes": null }, { "id": "2496303", "postDate": "10/24/2023 00:05:25", "content": "Are you talking about the RMDB? The data are freely available, no registration required. If you go to https://rmdb.stanford.edu/browse/, there's a sidebar on the left title \"Categories\" which includes a button to download selected files (by default, all the rdats published on the RMDB).
", "rawMarkdown": "Are you talking about the [RMDB](https://rmdb.stanford.edu/)? The data are freely available, no registration required. If you go to https://rmdb.stanford.edu/browse/, there's a sidebar on the left title \"Categories\" which includes a button to download selected files (by default, all the rdats published on the RMDB).", "votes": null }, { "id": "2496315", "postDate": "10/24/2023 00:43:40", "content": "Thanks - Yes I was talking about RMDB. I also tried the Eterna github but no luck
", "rawMarkdown": "Thanks - Yes I was talking about RMDB. I also tried the Eterna github but no luck", "votes": null }, { "id": "2503924", "postDate": "10/29/2023 14:46:47", "content": "Thanks Tomas! Do you by chance have pointers on how to process that data?
", "rawMarkdown": "Thanks Tomas! Do you by chance have pointers on how to process that data?", "votes": null } ], "comments": [ { "id": 2492980, "author_name": "callmeb", "author_url": "", "post_date": "10/23/2023 04:53:53", "content": "No free lunch theorem! Keep grinding :)
", "votes": null, "replies": [] }, { "id": 2493074, "author_name": "kirtirajsinhparmar", "author_url": "", "post_date": "10/23/2023 06:28:58", "content": "I tried to use conv but my training stuck at some point not improve i dont know why have you face this i wonder that if conv layers are effective or not here?
", "votes": null, "replies": [ { "id": 2493733, "author_name": "bonaventurefpdossou", "author_url": "", "post_date": "10/23/2023 14:41:50", "content": "Yes I faced the same issue
", "votes": null, "replies": [] } ] }, { "id": 2494015, "author_name": "digitalembrace", "author_url": "", "post_date": "10/23/2023 17:08:37", "content": "You could try pulling in more training data from RMDB or EternaBench or any other sources of RNA structure data you can find.
\nA really good model needs to understand 3D structure, which is a tall challenge. Has anyone explored AlphaFold style Evoformer modules?
", "votes": null, "replies": [ { "id": 2494024, "author_name": "bonaventurefpdossou", "author_url": "", "post_date": "10/23/2023 17:16:15", "content": "How to access the data? I signed up but could not get access to it. No,How to access the data? I signed up but could not get access to it. No I have not tried AF style models
", "votes": null, "replies": [ { "id": 2496303, "author_name": "brainbowrna", "author_url": "", "post_date": "10/24/2023 00:05:25", "content": "Are you talking about the RMDB? The data are freely available, no registration required. If you go to https://rmdb.stanford.edu/browse/, there's a sidebar on the left title \"Categories\" which includes a button to download selected files (by default, all the rdats published on the RMDB).
", "votes": null, "replies": [ { "id": 2496315, "author_name": "bonaventurefpdossou", "author_url": "", "post_date": "10/24/2023 00:43:40", "content": "Thanks - Yes I was talking about RMDB. I also tried the Eterna github but no luck
", "votes": null, "replies": [] }, { "id": 2503924, "author_name": "marianig82", "author_url": "", "post_date": "10/29/2023 14:46:47", "content": "Thanks Tomas! Do you by chance have pointers on how to process that data?
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "2492926": "Could you guys share and discuss some new ideas? (Without compromising the competition's and Kaggle's rules of course). I have been stuck for a couple of weeks now and none of my new ideas have been able to improve the performance - how can one use features like the signal to noise, etc?", "2492980": "No free lunch theorem! Keep grinding :)", "2493074": "I tried to use conv but my training stuck at some point not improve i dont know why have you face this i wonder that if conv layers are effective or not here?", "2493733": "Yes I faced the same issue", "2494015": "You could try pulling in more training data from [RMDB](https://rmdb.stanford.edu/) or [EternaBench](https://github.com/eternagame/EternaBench) or any other sources of RNA structure data you can find.\n\nA really good model needs to understand 3D structure, which is a tall challenge. Has anyone explored AlphaFold style Evoformer modules?", "2494024": "How to access the data? I signed up but could not get access to it. No,How to access the data? I signed up but could not get access to it. No I have not tried AF style models", "2496303": "Are you talking about the [RMDB](https://rmdb.stanford.edu/)? The data are freely available, no registration required. If you go to https://rmdb.stanford.edu/browse/, there's a sidebar on the left title \"Categories\" which includes a button to download selected files (by default, all the rdats published on the RMDB).", "2496315": "Thanks - Yes I was talking about RMDB. I also tried the Eterna github but no luck", "2503924": "Thanks Tomas! Do you by chance have pointers on how to process that data?" }, "source": "meta" }