We have only one submission per day here and, unfortunately, if an error occurs (which is not so unlikely given an unknown test set will be \"injected\") this will nevertheless count as such a submission! So let's wait until tomorrow… ;-)
", "messages": [ { "id": 1847344, "postDate": "2022-07-07T20:43:26.540Z", "content": "We have only one submission per day here and, unfortunately, if an error occurs (which is not so unlikely given an unknown test set will be \"injected\") this will nevertheless count as such a submission! So let's wait until tomorrow… ;-)
", "rawMarkdown": "We have only one submission per day here and, unfortunately, if an error occurs (which is not so unlikely given an unknown test set will be \"injected\") this will nevertheless count as such a submission! So let's wait until tomorrow... ;-)", "votes": 5 }, { "id": 1862585, "postDate": "2022-07-19T20:08:33.050Z", "content": "the reason is that there are a few test images and the metric is binary classification.
\none can probe the results easily
i would suggestion a new way of submitting for future competition like this :
\nor make another set of development/debug set for development (e.g. dummy image with same image size of real test set)
\nwhere don't use train set for debugging?
\nHmm, second day, second fail. Was anybody successful yet to submit something non-trivial?
\nMy try is here, but does not work… (submission scoring error):
\nhttps://www.kaggle.com/code/docxian/strip-ai-eda-of-structured-data
\nMore than happy, about any hints :-)
Finally found the problem: You need to remove duplicate entries in the submission stemming from the fact that patients can have more than one image => will occur multiple times in submission…
", "rawMarkdown": "Finally found the problem: You need to **remove duplicate entries** in the submission stemming from the fact that patients can have more than one image => will occur multiple times in submission...", "votes": 1 }, { "id": 1850953, "postDate": "2022-07-11T00:21:28.447Z", "content": "Failed 3x so far… @docxian you just dropped duplicates in test dataframe, or submission file?
", "rawMarkdown": "Failed 3x so far... @docxian you just dropped duplicates in test dataframe, or submission file?" }, { "id": 1851787, "postDate": "2022-07-11T14:39:44.903Z", "content": "I use the test set to build up a submission file. In the end I remove the duplicates.
\nSee here: https://www.kaggle.com/code/docxian/strip-ai-eda-of-structured-data
This is a very important point to note, one submission per day could be a huge challenge, especially at the end of the competition where each improved submission could determine the final standings.
", "rawMarkdown": "This is a very important point to note, one submission per day could be a huge challenge, especially at the end of the competition where each improved submission could determine the final standings." }, { "id": 1847510, "postDate": "2022-07-08T01:09:59.157Z", "content": "Thank you for saying this.. I nearly wasted it for nothing 😆
", "rawMarkdown": "Thank you for saying this.. I nearly wasted it for nothing 😆\n" } ], "comments": [ { "id": 1862585, "author_name": "hengck23", "author_url": "", "post_date": "2022-07-19T20:08:33.050000", "content": "the reason is that there are a few test images and the metric is binary classification.
\none can probe the results easily
i would suggestion a new way of submitting for future competition like this :
\nor make another set of development/debug set for development (e.g. dummy image with same image size of real test set)
\nwhere don't use train set for debugging?
\nHmm, second day, second fail. Was anybody successful yet to submit something non-trivial?
\nMy try is here, but does not work… (submission scoring error):
\nhttps://www.kaggle.com/code/docxian/strip-ai-eda-of-structured-data
\nMore than happy, about any hints :-)
Finally found the problem: You need to remove duplicate entries in the submission stemming from the fact that patients can have more than one image => will occur multiple times in submission…
", "votes": 1, "replies": [] }, { "id": 1850953, "author_name": "Tim Yee", "author_url": "", "post_date": "2022-07-11T00:21:28.447000", "content": "Failed 3x so far… @docxian you just dropped duplicates in test dataframe, or submission file?
", "votes": 0, "replies": [] }, { "id": 1851787, "author_name": "Chris X", "author_url": "", "post_date": "2022-07-11T14:39:44.903000", "content": "I use the test set to build up a submission file. In the end I remove the duplicates.
\nSee here: https://www.kaggle.com/code/docxian/strip-ai-eda-of-structured-data
This is a very important point to note, one submission per day could be a huge challenge, especially at the end of the competition where each improved submission could determine the final standings.
", "votes": 0, "replies": [] }, { "id": 1847510, "author_name": "The Devastator", "author_url": "", "post_date": "2022-07-08T01:09:59.157000", "content": "Thank you for saying this.. I nearly wasted it for nothing 😆
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1847344": "We have only one submission per day here and, unfortunately, if an error occurs (which is not so unlikely given an unknown test set will be \"injected\") this will nevertheless count as such a submission! So let's wait until tomorrow... ;-)", "1862585": "the reason is that there are a few test images and the metric is binary classification.\none can probe the results easily\n\n---\n\ni would suggestion a new way of submitting for future competition like this :\n- 5 submission per day (for development debugging) for the first few (e.g. 3) days you submitted\n- revert to 1 day per submission afterwards\n\nor make another set of development/debug set for development (e.g. dummy image with same image size of real test set)\n\nwhere don't use train set for debugging?\n- sometimes my notebook is ok using train set. but results in unknown error in submission. ", "1847829": "Hmm, second day, second fail. Was anybody successful yet to submit something non-trivial?\nMy try is here, but does not work... (submission scoring error):\nhttps://www.kaggle.com/code/docxian/strip-ai-eda-of-structured-data\nMore than happy, about any hints :-)", "1847775": "This is a very important point to note, one submission per day could be a huge challenge, especially at the end of the competition where each improved submission could determine the final standings.", "1847510": "Thank you for saying this.. I nearly wasted it for nothing 😆\n" } }