![\"\"](\"https://i.imgur.com/DTaTHdN.png\")
Just checked recently launched competition and found all Members have score 0.00000
\nEvaluation will be as per F1 score, so not getting why everyone getting lowest score !!!! ???
Just checked recently launched competition and found all Members have score 0.00000
\nEvaluation will be as per F1 score, so not getting why everyone getting lowest score !!!! ???
I don't think anyone has made actual predictions, most of the notebooks look like EDA
", "rawMarkdown": "I don't think anyone has made actual predictions, most of the notebooks look like EDA", "votes": 2, "replies": [ { "id": 1695354, "postDate": "2022-02-18T05:10:50.910Z", "content": "@outwrest , Yes .. Seems all had just made random submission.
\nBut hadn't expect that from public NB 😄
Another reason is there are a lot of classes and the metric is Macro-F1, so it will be pretty hard to step to the other score. Good luck, @kdsharmaai!
", "rawMarkdown": "Another reason is there are a lot of classes and the metric is Macro-F1, so it will be pretty hard to step to the other score. Good luck, @kdsharmaai! ", "votes": 1 }, { "id": 1695589, "postDate": "2022-02-18T08:35:03.970Z", "content": "@vad13irt , Yeah! That would be most probable reason
", "rawMarkdown": "@vad13irt , Yeah! That would be most probable reason ", "votes": 1 }, { "id": 1698260, "postDate": "2022-02-20T08:37:36.310Z", "content": "Hey, I have another reason! Is it because of overfitting?
", "rawMarkdown": "Hey, I have another reason! Is it because of overfitting?", "votes": 1 }, { "id": 1698386, "postDate": "2022-02-20T10:59:46.787Z", "content": "@dinowun I think that's unlikely, in that case we would have seen a much larger score 😄
", "rawMarkdown": "@dinowun I think that's unlikely, in that case we would have seen a much larger score 😄" }, { "id": 1698624, "postDate": "2022-02-20T14:39:31.313Z", "content": "@init27 not if everyone some manages to overfit to training dataset 😄😄. In reality, it looks like we have people with scores due to actual models. it is hard to work with all that data (160GB!!) so it makes sense that scores lag behind.
", "rawMarkdown": "@init27 not if everyone some manages to overfit to training dataset 😄😄. In reality, it looks like we have people with scores due to actual models. it is hard to work with all that data (160GB!!) so it makes sense that scores lag behind.", "votes": 1 }, { "id": 1698850, "postDate": "2022-02-20T18:01:29.173Z", "content": "Finally we can see some positive scoring 😄
", "rawMarkdown": "Finally we can see some positive scoring 😄" } ] }, { "id": 1707580, "postDate": "2022-02-28T15:35:07.440Z", "content": "No award ranking points + NO count towards tiers = No Motivation
", "rawMarkdown": "[No award ranking points + NO count towards tiers = No Motivation](https://www.kaggle.com/c/herbarium-2022-fgvc9/discussion/310277)" } ], "comments": [ { "id": 1694872, "author_name": "outwrest", "author_url": "", "post_date": "2022-02-17T19:30:02.627000", "content": "I don't think anyone has made actual predictions, most of the notebooks look like EDA
", "votes": 2, "replies": [ { "id": 1695354, "author_name": "Kiran Sharma", "author_url": "", "post_date": "2022-02-18T05:10:50.910000", "content": "@outwrest , Yes .. Seems all had just made random submission.
\nBut hadn't expect that from public NB 😄
Another reason is there are a lot of classes and the metric is Macro-F1, so it will be pretty hard to step to the other score. Good luck, @kdsharmaai!
", "votes": 1, "replies": [] }, { "id": 1695589, "author_name": "Kiran Sharma", "author_url": "", "post_date": "2022-02-18T08:35:03.970000", "content": "@vad13irt , Yeah! That would be most probable reason
", "votes": 1, "replies": [] }, { "id": 1698260, "author_name": "Dino Wun", "author_url": "", "post_date": "2022-02-20T08:37:36.310000", "content": "Hey, I have another reason! Is it because of overfitting?
", "votes": 1, "replies": [] }, { "id": 1698386, "author_name": "Sanyam Bhutani", "author_url": "", "post_date": "2022-02-20T10:59:46.787000", "content": "@dinowun I think that's unlikely, in that case we would have seen a much larger score 😄
", "votes": 0, "replies": [] }, { "id": 1698624, "author_name": "outwrest", "author_url": "", "post_date": "2022-02-20T14:39:31.313000", "content": "@init27 not if everyone some manages to overfit to training dataset 😄😄. In reality, it looks like we have people with scores due to actual models. it is hard to work with all that data (160GB!!) so it makes sense that scores lag behind.
", "votes": 1, "replies": [] }, { "id": 1698850, "author_name": "Kiran Sharma", "author_url": "", "post_date": "2022-02-20T18:01:29.173000", "content": "Finally we can see some positive scoring 😄
", "votes": 0, "replies": [] } ] }, { "id": 1707580, "author_name": "Faisal Alsrheed", "author_url": "", "post_date": "2022-02-28T15:35:07.440000", "content": "No award ranking points + NO count towards tiers = No Motivation
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1694482": "Just checked recently launched competition and found all Members have score 0.00000 \nEvaluation will be as per F1 score, so not getting why everyone getting lowest score !!!! ???\n\n", "1694872": "I don't think anyone has made actual predictions, most of the notebooks look like EDA", "1707580": "[No award ranking points + NO count towards tiers = No Motivation](https://www.kaggle.com/c/herbarium-2022-fgvc9/discussion/310277)" } }