It states that \"The private leaderboard is calculated over the same rows as the public leaderboard in this competition.\"\ndoes this mean we do not have to worry about the private leaderboard score because it is evaluated with same test set? Am I right?\nthanks and have a nice day!
", "messages": [ { "id": "519503", "postDate": "04/19/2019 04:52:30", "content": "It states that \"The private leaderboard is calculated over the same rows as the public leaderboard in this competition.\"\ndoes this mean we do not have to worry about the private leaderboard score because it is evaluated with same test set? Am I right?\nthanks and have a nice day!
", "rawMarkdown": "It states that \"The private leaderboard is calculated over the same rows as the public leaderboard in this competition.\"\ndoes this mean we do not have to worry about the private leaderboard score because it is evaluated with same test set? Am I right?\nthanks and have a nice day!", "votes": null }, { "id": "519547", "postDate": "04/19/2019 07:15:24", "content": "This stage - yes, same test set.
", "rawMarkdown": "This stage - yes, same test set.", "votes": null }, { "id": "519861", "postDate": "04/19/2019 18:39:59", "content": "No, this is a two stages competition. Actually, the private score will be evaluated on another test dataset, which we don't have access to (yet). That message on the leaderboard page is not correct. Carefully read the rules and the forum.
", "rawMarkdown": "No, this is a two stages competition. Actually, the private score will be evaluated on another test dataset, which we don't have access to (yet). That message on the leaderboard page is not correct. Carefully read the rules and the forum.", "votes": null }, { "id": "520079", "postDate": "04/20/2019 05:56:20", "content": "You can find the following statement at Data Description.
\n\n\n", "rawMarkdown": "You can find the following statement at [Data Description](https://www.kaggle.com/c/imet-2019-fgvc6/data).\n\n> This is a kernels-only competition **with two stages**. After the deadline, Kaggle will rerun your selected kernels **on an unseen test set**. The second-stage test set is approximately five times the size of the first. You should plan your kernel's memory, disk, and runtime footprint accordingly.", "votes": null }, { "id": "520087", "postDate": "04/20/2019 06:17:52", "content": "This is a kernels-only competition with two stages. After the deadline, Kaggle will rerun your selected kernels on an unseen test set. The second-stage test set is approximately five times the size of the first. You should plan your kernel's memory, disk, and runtime footprint accordingly.
\n
oh thanks I missed that point. so second stage is different to the first stage test set. Then how about the distribution of the test sets? could we say that the first stage test set and second stage test set has similar class distribution?
", "rawMarkdown": "oh thanks I missed that point. so second stage is different to the first stage test set. Then how about the distribution of the test sets? could we say that the first stage test set and second stage test set has similar class distribution?", "votes": null }, { "id": "520088", "postDate": "04/20/2019 06:18:38", "content": "thanks! unseen meaning we can't get any info about it right?
", "rawMarkdown": "thanks! unseen meaning we can't get any info about it right?", "votes": null }, { "id": "520101", "postDate": "04/20/2019 07:03:53", "content": "Yes, except for the size.
\n\n\n", "rawMarkdown": "Yes, except for the size.\n\n> The second-stage test set is approximately five times the size of the first.", "votes": null }, { "id": "520184", "postDate": "04/20/2019 11:22:40", "content": "The second-stage test set is approximately five times the size of the first.
\n
We don't know. I wouldn't rely on that.
", "rawMarkdown": "We don't know. I wouldn't rely on that.", "votes": null } ], "comments": [ { "id": 519547, "author_name": "konradb", "author_url": "", "post_date": "04/19/2019 07:15:24", "content": "This stage - yes, same test set.
", "votes": null, "replies": [] }, { "id": 519861, "author_name": "artyomp", "author_url": "", "post_date": "04/19/2019 18:39:59", "content": "No, this is a two stages competition. Actually, the private score will be evaluated on another test dataset, which we don't have access to (yet). That message on the leaderboard page is not correct. Carefully read the rules and the forum.
", "votes": null, "replies": [ { "id": 520087, "author_name": "yangsaewon", "author_url": "", "post_date": "04/20/2019 06:17:52", "content": "oh thanks I missed that point. so second stage is different to the first stage test set. Then how about the distribution of the test sets? could we say that the first stage test set and second stage test set has similar class distribution?
", "votes": null, "replies": [] }, { "id": 520184, "author_name": "artyomp", "author_url": "", "post_date": "04/20/2019 11:22:40", "content": "We don't know. I wouldn't rely on that.
", "votes": null, "replies": [] } ] }, { "id": 520079, "author_name": "ttahara", "author_url": "", "post_date": "04/20/2019 05:56:20", "content": "You can find the following statement at Data Description.
\n\n\n", "votes": null, "replies": [ { "id": 520088, "author_name": "yangsaewon", "author_url": "", "post_date": "04/20/2019 06:18:38", "content": "This is a kernels-only competition with two stages. After the deadline, Kaggle will rerun your selected kernels on an unseen test set. The second-stage test set is approximately five times the size of the first. You should plan your kernel's memory, disk, and runtime footprint accordingly.
\n
thanks! unseen meaning we can't get any info about it right?
", "votes": null, "replies": [] }, { "id": 520101, "author_name": "ttahara", "author_url": "", "post_date": "04/20/2019 07:03:53", "content": "Yes, except for the size.
\n\n\n", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "519503": "It states that \"The private leaderboard is calculated over the same rows as the public leaderboard in this competition.\"\ndoes this mean we do not have to worry about the private leaderboard score because it is evaluated with same test set? Am I right?\nthanks and have a nice day!", "519547": "This stage - yes, same test set.", "519861": "No, this is a two stages competition. Actually, the private score will be evaluated on another test dataset, which we don't have access to (yet). That message on the leaderboard page is not correct. Carefully read the rules and the forum.", "520079": "You can find the following statement at [Data Description](https://www.kaggle.com/c/imet-2019-fgvc6/data).\n\n> This is a kernels-only competition **with two stages**. After the deadline, Kaggle will rerun your selected kernels **on an unseen test set**. The second-stage test set is approximately five times the size of the first. You should plan your kernel's memory, disk, and runtime footprint accordingly.", "520087": "oh thanks I missed that point. so second stage is different to the first stage test set. Then how about the distribution of the test sets? could we say that the first stage test set and second stage test set has similar class distribution?", "520088": "thanks! unseen meaning we can't get any info about it right?", "520101": "Yes, except for the size.\n\n> The second-stage test set is approximately five times the size of the first.", "520184": "We don't know. I wouldn't rely on that." }, "source": "meta" }The second-stage test set is approximately five times the size of the first.
\n