![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F5eebaa7b9cd302745fc56833a450736e%2FSelection_999(8298).png?generation=1751328111063294&alt=media\")
that is why we have one shot and zero shot kagglers!
", "messages": [ { "id": "3237124", "postDate": "07/01/2025 00:02:08", "content": "
that is why we have one shot and zero shot kagglers!
", "rawMarkdown": ".png?generation=1751328111063294&alt=media)\n\nthat is why we have one shot and zero shot kagglers!", "votes": null }, { "id": "3237126", "postDate": "07/01/2025 00:03:33", "content": "Public and private scores are almost identicals.
", "rawMarkdown": "Public and private scores are almost identicals.", "votes": null }, { "id": "3237160", "postDate": "07/01/2025 00:26:20", "content": "Pretty rare!
\nLooking at my submissions, the private–public gap was always within 0.1—maybe that’s just because the dataset’s so huge.
One of the most unshaky competition ever.
\nIt's hardly shaky even when you go below 500.
It's probably because the dataset is large, according to the intro to probability theory that the law of large numbers and Central limit theorem, as the sample size is large, the sampling distribution would be almost the same as the population distribution!😀 And yes, it's almost the first time that I see this rigid ranking that have no shakeups! We'll see the neighboring JaneStreet Competition that if there would be shakeup there!
", "rawMarkdown": "It's probably because the dataset is large, according to the intro to probability theory that the law of large numbers and Central limit theorem, as the sample size is large, the sampling distribution would be almost the same as the population distribution!😀 And yes, it's almost the first time that I see this rigid ranking that have no shakeups! We'll see the neighboring JaneStreet Competition that if there would be shakeup there!", "votes": null }, { "id": "3237246", "postDate": "07/01/2025 01:56:49", "content": "This competition was actually incredible, some people only had 1 sub at that height too. I really wish I was in a position to do more for this competition because wow!
", "rawMarkdown": "This competition was actually incredible, some people only had 1 sub at that height too. I really wish I was in a position to do more for this competition because wow!", "votes": null }, { "id": "3237335", "postDate": "07/01/2025 03:27:57", "content": "there is no shakeup NOT becuase the train data is large.
\nIt is because you are supposed to use test data in your model building.
Top solutions either use it as \"pesudo label\" in bulk or use overfit network one by one,
", "rawMarkdown": "there is no shakeup NOT becuase the train data is large.\nIt is because you are supposed to use test data in your model building.\n\nTop solutions either use it as \"pesudo label\" in bulk or use overfit network one by one,", "votes": null }, { "id": "3237365", "postDate": "07/01/2025 03:53:10", "content": "Thanks for clarifying! I think that is the root cause of it.
", "rawMarkdown": "Thanks for clarifying! I think that is the root cause of it.", "votes": null }, { "id": "3237463", "postDate": "07/01/2025 05:41:45", "content": "Not necessarily I didn't. (But it certainly explain why many Kagglers got better result than I expected).
", "rawMarkdown": "Not necessarily I didn't. (But it certainly explain why many Kagglers got better result than I expected).", "votes": null }, { "id": "3237742", "postDate": "07/01/2025 09:19:09", "content": "Reminds me of a quote from Thanos from Infinity War \"Perfectly balance as things should be\".
", "rawMarkdown": "Reminds me of a quote from Thanos from Infinity War \"Perfectly balance as things should be\".", "votes": null } ], "comments": [ { "id": 3237126, "author_name": "sacuscreed", "author_url": "", "post_date": "07/01/2025 00:03:33", "content": "Public and private scores are almost identicals.
", "votes": null, "replies": [] }, { "id": 3237160, "author_name": "iwaiwa", "author_url": "", "post_date": "07/01/2025 00:26:20", "content": "Pretty rare!
\nLooking at my submissions, the private–public gap was always within 0.1—maybe that’s just because the dataset’s so huge.
One of the most unshaky competition ever.
\nIt's hardly shaky even when you go below 500.
It's probably because the dataset is large, according to the intro to probability theory that the law of large numbers and Central limit theorem, as the sample size is large, the sampling distribution would be almost the same as the population distribution!😀 And yes, it's almost the first time that I see this rigid ranking that have no shakeups! We'll see the neighboring JaneStreet Competition that if there would be shakeup there!
", "votes": null, "replies": [] }, { "id": 3237246, "author_name": "cody11null", "author_url": "", "post_date": "07/01/2025 01:56:49", "content": "This competition was actually incredible, some people only had 1 sub at that height too. I really wish I was in a position to do more for this competition because wow!
", "votes": null, "replies": [] }, { "id": 3237335, "author_name": "hengck23", "author_url": "", "post_date": "07/01/2025 03:27:57", "content": "there is no shakeup NOT becuase the train data is large.
\nIt is because you are supposed to use test data in your model building.
Top solutions either use it as \"pesudo label\" in bulk or use overfit network one by one,
", "votes": null, "replies": [ { "id": 3237365, "author_name": "huasteven", "author_url": "", "post_date": "07/01/2025 03:53:10", "content": "Thanks for clarifying! I think that is the root cause of it.
", "votes": null, "replies": [] }, { "id": 3237463, "author_name": "shlomoron", "author_url": "", "post_date": "07/01/2025 05:41:45", "content": "Not necessarily I didn't. (But it certainly explain why many Kagglers got better result than I expected).
", "votes": null, "replies": [] } ] }, { "id": 3237742, "author_name": "sayedathar11", "author_url": "", "post_date": "07/01/2025 09:19:09", "content": "Reminds me of a quote from Thanos from Infinity War \"Perfectly balance as things should be\".
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3237124": ".png?generation=1751328111063294&alt=media)\n\nthat is why we have one shot and zero shot kagglers!", "3237126": "Public and private scores are almost identicals.", "3237160": "Pretty rare! \nLooking at my submissions, the private–public gap was always within 0.1—maybe that’s just because the dataset’s so huge.", "3237170": "One of the most unshaky competition ever.\nIt's hardly shaky even when you go below 500.", "3237185": "It's probably because the dataset is large, according to the intro to probability theory that the law of large numbers and Central limit theorem, as the sample size is large, the sampling distribution would be almost the same as the population distribution!😀 And yes, it's almost the first time that I see this rigid ranking that have no shakeups! We'll see the neighboring JaneStreet Competition that if there would be shakeup there!", "3237246": "This competition was actually incredible, some people only had 1 sub at that height too. I really wish I was in a position to do more for this competition because wow!", "3237335": "there is no shakeup NOT becuase the train data is large.\nIt is because you are supposed to use test data in your model building.\n\nTop solutions either use it as \"pesudo label\" in bulk or use overfit network one by one,", "3237365": "Thanks for clarifying! I think that is the root cause of it.", "3237463": "Not necessarily I didn't. (But it certainly explain why many Kagglers got better result than I expected).", "3237742": "Reminds me of a quote from Thanos from Infinity War \"Perfectly balance as things should be\"." }, "source": "meta" }