Since the current LB looks so different from the Stage-1 LB, despite the fact that we know it is based on 1% or approximately 30 images, there are concerns.
\n\nSo I am sharing a pdf of the stage-1 LB that I saved for my reference. Mind you it is not the whole thing but the type you see after making submission i.e. a jump in the middle to show my score and then up to only 247th team.
\n\nMy take on the current LB is really not much but to use as a reference. I made a whole bunch of submissions to the stage-1 LB for models that finished training after the close of Stage-1 so that I can compare my local CV performance. Even though a couple of my models scored slightly higher than my chosen submission, they all had a higher CV score than my final choice on the stage-1 LB. I made those submissions before the stage-1 LB was wiped out. I observed similar difference between my CV and LB scores on the current (i.e. stage-2) not so useful public LB.
\n\nEnjoy.
", "messages": [ { "id": "412207", "postDate": "10/29/2018 19:32:31", "content": "Since the current LB looks so different from the Stage-1 LB, despite the fact that we know it is based on 1% or approximately 30 images, there are concerns.
\n\nSo I am sharing a pdf of the stage-1 LB that I saved for my reference. Mind you it is not the whole thing but the type you see after making submission i.e. a jump in the middle to show my score and then up to only 247th team.
\n\nMy take on the current LB is really not much but to use as a reference. I made a whole bunch of submissions to the stage-1 LB for models that finished training after the close of Stage-1 so that I can compare my local CV performance. Even though a couple of my models scored slightly higher than my chosen submission, they all had a higher CV score than my final choice on the stage-1 LB. I made those submissions before the stage-1 LB was wiped out. I observed similar difference between my CV and LB scores on the current (i.e. stage-2) not so useful public LB.
\n\nEnjoy.
", "rawMarkdown": "Since the current LB looks so different from the Stage-1 LB, despite the fact that we know it is based on 1% or approximately 30 images, there are concerns.\n\nSo I am sharing a [pdf of the stage-1 LB][1] that I saved for my reference. Mind you it is not the whole thing but the type you see after making submission i.e. a jump in the middle to show my score and then up to only 247th team.\n\nMy take on the current LB is really not much but to use as a reference. I made a whole bunch of submissions to the stage-1 LB for models that finished training after the close of Stage-1 so that I can compare my local CV performance. Even though a couple of my models scored slightly higher than my chosen submission, they all had a higher CV score than my final choice on the stage-1 LB. I made those submissions before the stage-1 LB was wiped out. I observed similar difference between my CV and LB scores on the current (i.e. stage-2) not so useful public LB.\n\nEnjoy.\n\n\n [1]: https://www.kaggle.com/sheriytm/get-yagana-s-stage1-2-pdf-kaggle-bug-workaround/output", "votes": null }, { "id": "412212", "postDate": "10/29/2018 19:37:57", "content": "Thanks, YaGana. It looks like it's not online, though. 404 - We can't find that page.
", "rawMarkdown": "Thanks, YaGana. It looks like it's not online, though. 404 - We can't find that page.", "votes": null }, { "id": "412216", "postDate": "10/29/2018 19:45:12", "content": "Fixed. I mistakenly attached my private data set before. It should work now. Please let me know if you are able to download the pdf file.
", "rawMarkdown": "Fixed. I mistakenly attached my private data set before. It should work now. Please let me know if you are able to download the pdf file.", "votes": null }, { "id": "412852", "postDate": "10/30/2018 22:39:24", "content": "I just updated the last paragraph in my post to make it clearer. I submitted the results of my late models to the stage-1 LB 1st before it was wiped out. I then submitted those to the current i.e. stage-2 LB to compare.
", "rawMarkdown": "I just updated the last paragraph in my post to make it clearer. I submitted the results of my late models to the stage-1 LB 1st before it was wiped out. I then submitted those to the current i.e. stage-2 LB to compare.", "votes": null }, { "id": "413441", "postDate": "11/01/2018 00:48:55", "content": "I still can't get it to download directly, but kernels can read it, so I wrote a kernel to just copy it. I can download it from the kernel output. Hopefully others can too.
", "rawMarkdown": "I still can't get it to download directly, but kernels can read it, so I wrote [a kernel to just copy it][1]. I can download it from the kernel output. Hopefully others can too.\n\n [1]: https://www.kaggle.com/aharless/get-yagana-s-stage1-pdf-kaggle-bug-workaround", "votes": null }, { "id": "413478", "postDate": "11/01/2018 02:46:05", "content": "Thanks @Andy Harless. I may be in the middle of uploading the Stage-2 public LB pdf when you tried. Kaggle froze on me 3 times before I was able to upload that one. It also froze when I was creating the discussion post and ended up posting that twice. Anyway the stage-2 public LB is here https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/70216
", "rawMarkdown": "Thanks @Andy Harless. I may be in the middle of uploading the Stage-2 public LB pdf when you tried. Kaggle froze on me 3 times before I was able to upload that one. It also froze when I was creating the discussion post and ended up posting that twice. Anyway the stage-2 public LB is here https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/70216", "votes": null } ], "comments": [ { "id": 412212, "author_name": "felipekitamura", "author_url": "", "post_date": "10/29/2018 19:37:57", "content": "Thanks, YaGana. It looks like it's not online, though. 404 - We can't find that page.
", "votes": null, "replies": [ { "id": 412216, "author_name": "sheriytm", "author_url": "", "post_date": "10/29/2018 19:45:12", "content": "Fixed. I mistakenly attached my private data set before. It should work now. Please let me know if you are able to download the pdf file.
", "votes": null, "replies": [] } ] }, { "id": 412852, "author_name": "sheriytm", "author_url": "", "post_date": "10/30/2018 22:39:24", "content": "I just updated the last paragraph in my post to make it clearer. I submitted the results of my late models to the stage-1 LB 1st before it was wiped out. I then submitted those to the current i.e. stage-2 LB to compare.
", "votes": null, "replies": [] }, { "id": 413441, "author_name": "aharless", "author_url": "", "post_date": "11/01/2018 00:48:55", "content": "I still can't get it to download directly, but kernels can read it, so I wrote a kernel to just copy it. I can download it from the kernel output. Hopefully others can too.
", "votes": null, "replies": [ { "id": 413478, "author_name": "sheriytm", "author_url": "", "post_date": "11/01/2018 02:46:05", "content": "Thanks @Andy Harless. I may be in the middle of uploading the Stage-2 public LB pdf when you tried. Kaggle froze on me 3 times before I was able to upload that one. It also froze when I was creating the discussion post and ended up posting that twice. Anyway the stage-2 public LB is here https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/70216
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "412207": "Since the current LB looks so different from the Stage-1 LB, despite the fact that we know it is based on 1% or approximately 30 images, there are concerns.\n\nSo I am sharing a [pdf of the stage-1 LB][1] that I saved for my reference. Mind you it is not the whole thing but the type you see after making submission i.e. a jump in the middle to show my score and then up to only 247th team.\n\nMy take on the current LB is really not much but to use as a reference. I made a whole bunch of submissions to the stage-1 LB for models that finished training after the close of Stage-1 so that I can compare my local CV performance. Even though a couple of my models scored slightly higher than my chosen submission, they all had a higher CV score than my final choice on the stage-1 LB. I made those submissions before the stage-1 LB was wiped out. I observed similar difference between my CV and LB scores on the current (i.e. stage-2) not so useful public LB.\n\nEnjoy.\n\n\n [1]: https://www.kaggle.com/sheriytm/get-yagana-s-stage1-2-pdf-kaggle-bug-workaround/output", "412212": "Thanks, YaGana. It looks like it's not online, though. 404 - We can't find that page.", "412216": "Fixed. I mistakenly attached my private data set before. It should work now. Please let me know if you are able to download the pdf file.", "412852": "I just updated the last paragraph in my post to make it clearer. I submitted the results of my late models to the stage-1 LB 1st before it was wiped out. I then submitted those to the current i.e. stage-2 LB to compare.", "413441": "I still can't get it to download directly, but kernels can read it, so I wrote [a kernel to just copy it][1]. I can download it from the kernel output. Hopefully others can too.\n\n [1]: https://www.kaggle.com/aharless/get-yagana-s-stage1-pdf-kaggle-bug-workaround", "413478": "Thanks @Andy Harless. I may be in the middle of uploading the Stage-2 public LB pdf when you tried. Kaggle froze on me 3 times before I was able to upload that one. It also froze when I was creating the discussion post and ended up posting that twice. Anyway the stage-2 public LB is here https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/70216" }, "source": "meta" }