I am really confused that what exactly is going on. The stage 2 submission file includes stage 1 test set as well, in the meantime the ground truth values for the stage 1 test set has been released. All the Kagglers uploaded 0.00000 submission.
\n\nMy Question: What should we do, do we have to submit our own prediction through the model which we submitted or add into submission file the ground truth values for first 512 test images?
\n\nMoreover, one Kaggler found leak images in stage 2 test data which is totally against the rule as ADMINS mentioned earlier that there would be totally new images. If the leak still remains the people who made LB mining get advantages of this fact.
\n\nFinal point, are we really want to help to treat cancer or just playing around to get the prizes at the end? It was so much fun to compete but when there is so much new issues which reduce the competition's interest is kinda not good.
", "messages": [ { "id": "193168", "postDate": "06/15/2017 18:46:45", "content": "I am really confused that what exactly is going on. The stage 2 submission file includes stage 1 test set as well, in the meantime the ground truth values for the stage 1 test set has been released. All the Kagglers uploaded 0.00000 submission.
\n\nMy Question: What should we do, do we have to submit our own prediction through the model which we submitted or add into submission file the ground truth values for first 512 test images?
\n\nMoreover, one Kaggler found leak images in stage 2 test data which is totally against the rule as ADMINS mentioned earlier that there would be totally new images. If the leak still remains the people who made LB mining get advantages of this fact.
\n\nFinal point, are we really want to help to treat cancer or just playing around to get the prizes at the end? It was so much fun to compete but when there is so much new issues which reduce the competition's interest is kinda not good.
", "rawMarkdown": "I am really confused that what exactly is going on. The stage 2 submission file includes stage 1 test set as well, in the meantime the ground truth values for the stage 1 test set has been released. All the Kagglers uploaded 0.00000 submission. \n\nMy Question: What should we do, do we have to submit our own prediction through the model which we submitted or add into submission file the ground truth values for first 512 test images?\n\nMoreover, one Kaggler found leak images in stage 2 test data which is totally against the rule as ADMINS mentioned earlier that there would be totally new images. If the leak still remains the people who made LB mining get advantages of this fact. \n\nFinal point, are we really want to help to treat cancer or just playing around to get the prizes at the end? It was so much fun to compete but when there is so much new issues which reduce the competition's interest is kinda not good.", "votes": null }, { "id": "193172", "postDate": "06/15/2017 18:57:53", "content": "While I agree with you that the recently discovered leak is a big deal, the leak appears to involve images from the original training dataset as well(but it still definitely benefits the public LB miners more). In terms of the people getting a 0.00000 submission on the public leaderboard, that will be completely useless on the final rankings. The public leaderboard is 13% of the overall test data but the final private rankings will only be scored on the remaining unseen 87%(hopefully besides the leak images). Including the stage 1 ground truth values will have no effect on the final standings.
", "rawMarkdown": "While I agree with you that the recently discovered leak is a big deal, the leak appears to involve images from the original training dataset as well(but it still definitely benefits the public LB miners more). In terms of the people getting a 0.00000 submission on the public leaderboard, that will be completely useless on the final rankings. The public leaderboard is 13% of the overall test data but the final private rankings will only be scored on the remaining unseen 87%(hopefully besides the leak images). Including the stage 1 ground truth values will have no effect on the final standings.", "votes": null }, { "id": "193179", "postDate": "06/15/2017 19:17:28", "content": "I hope so @timothy, it goes same as you mentioned.
", "rawMarkdown": "I hope so @timothy, it goes same as you mentioned.", "votes": null }, { "id": "193504", "postDate": "06/16/2017 18:06:16", "content": "About the similar images, please see my response in this thread.
", "rawMarkdown": "About the similar images, please see my response in [this thread][1]. \n\n\n [1]: https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/34788#193233", "votes": null } ], "comments": [ { "id": 193172, "author_name": "timothyrimnac", "author_url": "", "post_date": "06/15/2017 18:57:53", "content": "While I agree with you that the recently discovered leak is a big deal, the leak appears to involve images from the original training dataset as well(but it still definitely benefits the public LB miners more). In terms of the people getting a 0.00000 submission on the public leaderboard, that will be completely useless on the final rankings. The public leaderboard is 13% of the overall test data but the final private rankings will only be scored on the remaining unseen 87%(hopefully besides the leak images). Including the stage 1 ground truth values will have no effect on the final standings.
", "votes": null, "replies": [ { "id": 193179, "author_name": "svesal", "author_url": "", "post_date": "06/15/2017 19:17:28", "content": "I hope so @timothy, it goes same as you mentioned.
", "votes": null, "replies": [] } ] }, { "id": 193504, "author_name": "wendykan", "author_url": "", "post_date": "06/16/2017 18:06:16", "content": "About the similar images, please see my response in this thread.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "193168": "I am really confused that what exactly is going on. The stage 2 submission file includes stage 1 test set as well, in the meantime the ground truth values for the stage 1 test set has been released. All the Kagglers uploaded 0.00000 submission. \n\nMy Question: What should we do, do we have to submit our own prediction through the model which we submitted or add into submission file the ground truth values for first 512 test images?\n\nMoreover, one Kaggler found leak images in stage 2 test data which is totally against the rule as ADMINS mentioned earlier that there would be totally new images. If the leak still remains the people who made LB mining get advantages of this fact. \n\nFinal point, are we really want to help to treat cancer or just playing around to get the prizes at the end? It was so much fun to compete but when there is so much new issues which reduce the competition's interest is kinda not good.", "193172": "While I agree with you that the recently discovered leak is a big deal, the leak appears to involve images from the original training dataset as well(but it still definitely benefits the public LB miners more). In terms of the people getting a 0.00000 submission on the public leaderboard, that will be completely useless on the final rankings. The public leaderboard is 13% of the overall test data but the final private rankings will only be scored on the remaining unseen 87%(hopefully besides the leak images). Including the stage 1 ground truth values will have no effect on the final standings.", "193179": "I hope so @timothy, it goes same as you mentioned.", "193504": "About the similar images, please see my response in [this thread][1]. \n\n\n [1]: https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/34788#193233" }, "source": "meta" }