I learned from this website: https://www.kaggle.com/competitions/lish-moa/discussion/202256 that TabNet and 1dCNN also have a good effect on bioinformatics prediction….but I don't have time to reproduce this methodology since this is my first time to attend a kaggle competition…but I am wondering whether there are people trying this, and get good results? I think the methodology here is very attractive, especially the idea adding an FC layer before using 1d CNN…I just tried MLP. Now my codes of TabNet and Conv1D are still running…no chance to get results out before the ddl….(I am using the server from our university which has needs to load their module of tensorflow if you want to use gpu…which leads to a conflict…so I was running all my codes with cpu….
", "messages": [ { "id": 2031160, "postDate": "2022-11-15T22:51:33.797Z", "content": "I learned from this website: https://www.kaggle.com/competitions/lish-moa/discussion/202256 that TabNet and 1dCNN also have a good effect on bioinformatics prediction….but I don't have time to reproduce this methodology since this is my first time to attend a kaggle competition…but I am wondering whether there are people trying this, and get good results? I think the methodology here is very attractive, especially the idea adding an FC layer before using 1d CNN…I just tried MLP. Now my codes of TabNet and Conv1D are still running…no chance to get results out before the ddl….(I am using the server from our university which has needs to load their module of tensorflow if you want to use gpu…which leads to a conflict…so I was running all my codes with cpu….
", "rawMarkdown": "I learned from this website: https://www.kaggle.com/competitions/lish-moa/discussion/202256 that TabNet and 1dCNN also have a good effect on bioinformatics prediction....but I don't have time to reproduce this methodology since this is my first time to attend a kaggle competition...but I am wondering whether there are people trying this, and get good results? I think the methodology here is very attractive, especially the idea adding an FC layer before using 1d CNN...I just tried MLP. Now my codes of TabNet and Conv1D are still running...no chance to get results out before the ddl....(I am using the server from our university which has needs to load their module of tensorflow if you want to use gpu...which leads to a conflict...so I was running all my codes with cpu....", "votes": 1 }, { "id": 2033910, "postDate": "2022-11-17T16:44:49.183Z", "content": "Hey bro I really do not feel this post should be criticized for those reasons, with soooooo many other ideas posted in the discussion area, this is not code, not submission, no high score guaranteed. Cheer up, I support you. 🙏
", "rawMarkdown": "Hey bro I really do not feel this post should be criticized for those reasons, with soooooo many other ideas posted in the discussion area, this is not code, not submission, no high score guaranteed. Cheer up, I support you. 🙏", "votes": 1 }, { "id": 2031623, "postDate": "2022-11-16T07:24:15.230Z", "content": "Our team tried 1dCNN, but its initial results were so poor that we didn't explore it further.
", "rawMarkdown": "Our team tried 1dCNN, but its initial results were so poor that we didn't explore it further.", "votes": 1 }, { "id": 2031226, "postDate": "2022-11-16T00:23:25.587Z", "content": "Hey man! I did use Id CNN and it worked flawlessly!!!!!!
", "rawMarkdown": "Hey man! I did use Id CNN and it worked flawlessly!!!!!!", "votes": 1, "replies": [ { "id": 2041357, "postDate": "2022-11-23T21:09:57.367Z", "content": "Thanks! It's a pity that I didn't have time to submit results of CNN….but glad to see it worked! Maybe next year I can start with it.
", "rawMarkdown": "Thanks! It's a pity that I didn't have time to submit results of CNN....but glad to see it worked! Maybe next year I can start with it." } ] }, { "id": 2031166, "postDate": "2022-11-15T22:59:01.217Z", "content": "I assumed you were talking about multiome task. I tried the transfomer encoder like idea but couldn't come up with anything better than a simple MLP.
\nlet me know if you get anything positive and we could talk more about this
I was just wondering whether an idea works or not…but why did @callmeb posted a reply like this? ![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12066256%2F529fed3d681ce63737aff451d1c4d346%2FC72EE257-5AD8-4AB7-8908-0B7C222E1F10.jpeg?generation=1668555107198562&alt=media\")
\nI clearly stated that it’s my first time to attend Kaggle competition…also one can track my discussions under other posts. Truly I was met with guys unfriendly like this one who also clicked a downvote for my post. Hopefully Kaggle will launch a course of rules for communication and ethics. High scores on competitions does not mean a high score on personality. lol
Ok ok, maybe I was not clear but I did not accuse you to cheat.
\nIn many competitions happen that people post content like this few hours before the end of the comp (and someone can exploit it) This is not good!
\nYou could have just wait for 2h! And everything would been fine. Just let you know this.
I’m a really easy going guy, we can have a chat whenever you want I will be happy to talk with you :)
", "rawMarkdown": "Ok ok, maybe I was not clear but I did not accuse you to cheat.\n\nIn many competitions happen that people post content like this few hours before the end of the comp (and someone can exploit it) This is not good!\nYou could have just wait for 2h! And everything would been fine. Just let you know this.\n\nI’m a really easy going guy, we can have a chat whenever you want I will be happy to talk with you :)" } ] }, { "id": 2031254, "postDate": "2022-11-16T00:57:09.220Z", "content": "I tried 1D-CNN, but the correlation of the results was highly similar to MLP, so I gave it up.
", "rawMarkdown": "I tried 1D-CNN, but the correlation of the results was highly similar to MLP, so I gave it up." }, { "id": 2031383, "postDate": "2022-11-16T03:35:26.227Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2031173, "postDate": "2022-11-15T23:07:03.627Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2033910, "author_name": "jinyang18", "author_url": "", "post_date": "2022-11-17T16:44:49.183000", "content": "Hey bro I really do not feel this post should be criticized for those reasons, with soooooo many other ideas posted in the discussion area, this is not code, not submission, no high score guaranteed. Cheer up, I support you. 🙏
", "votes": 1, "replies": [] }, { "id": 2031623, "author_name": "Leo", "author_url": "", "post_date": "2022-11-16T07:24:15.230000", "content": "Our team tried 1dCNN, but its initial results were so poor that we didn't explore it further.
", "votes": 1, "replies": [] }, { "id": 2031226, "author_name": "SergioMiguelM", "author_url": "", "post_date": "2022-11-16T00:23:25.587000", "content": "Hey man! I did use Id CNN and it worked flawlessly!!!!!!
", "votes": 1, "replies": [ { "id": 2041357, "author_name": "XuShanKaggle", "author_url": "", "post_date": "2022-11-23T21:09:57.367000", "content": "Thanks! It's a pity that I didn't have time to submit results of CNN….but glad to see it worked! Maybe next year I can start with it.
", "votes": 0, "replies": [] } ] }, { "id": 2031166, "author_name": "yjyang", "author_url": "", "post_date": "2022-11-15T22:59:01.217000", "content": "I assumed you were talking about multiome task. I tried the transfomer encoder like idea but couldn't come up with anything better than a simple MLP.
\nlet me know if you get anything positive and we could talk more about this
I was just wondering whether an idea works or not…but why did @callmeb posted a reply like this?
\nI clearly stated that it’s my first time to attend Kaggle competition…also one can track my discussions under other posts. Truly I was met with guys unfriendly like this one who also clicked a downvote for my post. Hopefully Kaggle will launch a course of rules for communication and ethics. High scores on competitions does not mean a high score on personality. lol
Ok ok, maybe I was not clear but I did not accuse you to cheat.
\nIn many competitions happen that people post content like this few hours before the end of the comp (and someone can exploit it) This is not good!
\nYou could have just wait for 2h! And everything would been fine. Just let you know this.
I’m a really easy going guy, we can have a chat whenever you want I will be happy to talk with you :)
", "votes": 0, "replies": [] } ] }, { "id": 2031254, "author_name": "A Beginner", "author_url": "", "post_date": "2022-11-16T00:57:09.220000", "content": "I tried 1D-CNN, but the correlation of the results was highly similar to MLP, so I gave it up.
", "votes": 0, "replies": [] }, { "id": 2031383, "author_name": "", "author_url": "", "post_date": "2022-11-16T03:35:26.227000", "content": "", "votes": 0, "replies": [] }, { "id": 2031173, "author_name": "", "author_url": "", "post_date": "2022-11-15T23:07:03.627000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2031160": "I learned from this website: https://www.kaggle.com/competitions/lish-moa/discussion/202256 that TabNet and 1dCNN also have a good effect on bioinformatics prediction....but I don't have time to reproduce this methodology since this is my first time to attend a kaggle competition...but I am wondering whether there are people trying this, and get good results? I think the methodology here is very attractive, especially the idea adding an FC layer before using 1d CNN...I just tried MLP. Now my codes of TabNet and Conv1D are still running...no chance to get results out before the ddl....(I am using the server from our university which has needs to load their module of tensorflow if you want to use gpu...which leads to a conflict...so I was running all my codes with cpu....", "2033910": "Hey bro I really do not feel this post should be criticized for those reasons, with soooooo many other ideas posted in the discussion area, this is not code, not submission, no high score guaranteed. Cheer up, I support you. 🙏", "2031623": "Our team tried 1dCNN, but its initial results were so poor that we didn't explore it further.", "2031226": "Hey man! I did use Id CNN and it worked flawlessly!!!!!!", "2031166": "I assumed you were talking about multiome task. I tried the transfomer encoder like idea but couldn't come up with anything better than a simple MLP. \nlet me know if you get anything positive and we could talk more about this", "2031196": "I was just wondering whether an idea works or not…but why did @callmeb posted a reply like this? \nI clearly stated that it’s my first time to attend Kaggle competition…also one can track my discussions under other posts. Truly I was met with guys unfriendly like this one who also clicked a downvote for my post. Hopefully Kaggle will launch a course of rules for communication and ethics. High scores on competitions does not mean a high score on personality. lol", "2031254": "I tried 1D-CNN, but the correlation of the results was highly similar to MLP, so I gave it up.", "2031383": "", "2031173": "" } }