This is my first Kaggle competition so I want to know if the drastic changes between the public and private leader board are normal for a kaggle competition. I'm thinking that the change may be abnormal due to the number of outliers.
", "messages": [ { "id": "479070", "postDate": "02/27/2019 00:41:51", "content": "This is my first Kaggle competition so I want to know if the drastic changes between the public and private leader board are normal for a kaggle competition. I'm thinking that the change may be abnormal due to the number of outliers.
", "rawMarkdown": "This is my first Kaggle competition so I want to know if the drastic changes between the public and private leader board are normal for a kaggle competition. I'm thinking that the change may be abnormal due to the number of outliers.", "votes": null }, { "id": "479072", "postDate": "02/27/2019 00:43:17", "content": "There's a Shake up! Some top teams disappeared.
", "rawMarkdown": "There's a Shake up! Some top teams disappeared.", "votes": null }, { "id": "479095", "postDate": "02/27/2019 00:56:59", "content": "That depends on the distribution of data in the train and test(hidden) data. \nif they have the same distribution, all goes well.\notherwise, there will be a big change between the public and private LB.
\n\nThat's why Kaggle permit that you select two kernels for Your final score.\nSo select one that gives a good public score ie where you trust the public LB\nand another one that gives a good CV score because the CV score is more reliable if data changes.
", "rawMarkdown": "That depends on the distribution of data in the train and test(hidden) data. \nif they have the same distribution, all goes well.\notherwise, there will be a big change between the public and private LB.\n\nThat's why Kaggle permit that you select two kernels for Your final score.\nSo select one that gives a good public score ie where you trust the public LB\nand another one that gives a good CV score because the CV score is more reliable if data changes.", "votes": null }, { "id": "479107", "postDate": "02/27/2019 01:08:46", "content": "Yeah, there were too much shake ups. I wonder how some of the top teams at the public leaderboard managed to stay on top in the private leaderboard. Looking forward for their strategies that made robust submissions.
", "rawMarkdown": "Yeah, there were too much shake ups. I wonder how some of the top teams at the public leaderboard managed to stay on top in the private leaderboard. Looking forward for their strategies that made robust submissions.", "votes": null }, { "id": "479115", "postDate": "02/27/2019 01:17:00", "content": "It so happened that I didn't really trust CV, for which I was punished.
", "rawMarkdown": "It so happened that I didn't really trust CV, for which I was punished.", "votes": null }, { "id": "479146", "postDate": "02/27/2019 01:41:07", "content": "So big gap with other teams, you make a amazing work.
", "rawMarkdown": "So big gap with other teams, you make a amazing work.", "votes": null }, { "id": "479155", "postDate": "02/27/2019 01:49:34", "content": "The outliers were really difficult to deal with, but, anyway, you did a great work!
", "rawMarkdown": "The outliers were really difficult to deal with, but, anyway, you did a great work!", "votes": null }, { "id": "479485", "postDate": "02/27/2019 07:28:43", "content": "Congratulations @petersk20 for getting your first medal.
", "rawMarkdown": "Congratulations @petersk20 for getting your first medal.", "votes": null }, { "id": "479489", "postDate": "02/27/2019 07:31:20", "content": "You did really an amazing work @jingobox. You deserve a medal, always it is the best option to select two submissions with high CV and high LB.
", "rawMarkdown": "You did really an amazing work @jingobox. You deserve a medal, always it is the best option to select two submissions with high CV and high LB.", "votes": null } ], "comments": [ { "id": 479072, "author_name": "sreeedevi", "author_url": "", "post_date": "02/27/2019 00:43:17", "content": "There's a Shake up! Some top teams disappeared.
", "votes": null, "replies": [] }, { "id": 479095, "author_name": "jmourad100", "author_url": "", "post_date": "02/27/2019 00:56:59", "content": "That depends on the distribution of data in the train and test(hidden) data. \nif they have the same distribution, all goes well.\notherwise, there will be a big change between the public and private LB.
\n\nThat's why Kaggle permit that you select two kernels for Your final score.\nSo select one that gives a good public score ie where you trust the public LB\nand another one that gives a good CV score because the CV score is more reliable if data changes.
", "votes": null, "replies": [] }, { "id": 479107, "author_name": "harangdev", "author_url": "", "post_date": "02/27/2019 01:08:46", "content": "Yeah, there were too much shake ups. I wonder how some of the top teams at the public leaderboard managed to stay on top in the private leaderboard. Looking forward for their strategies that made robust submissions.
", "votes": null, "replies": [] }, { "id": 479115, "author_name": "jingobox", "author_url": "", "post_date": "02/27/2019 01:17:00", "content": "It so happened that I didn't really trust CV, for which I was punished.
", "votes": null, "replies": [ { "id": 479146, "author_name": "codlife", "author_url": "", "post_date": "02/27/2019 01:41:07", "content": "So big gap with other teams, you make a amazing work.
", "votes": null, "replies": [] }, { "id": 479155, "author_name": "rmazzine", "author_url": "", "post_date": "02/27/2019 01:49:34", "content": "The outliers were really difficult to deal with, but, anyway, you did a great work!
", "votes": null, "replies": [] }, { "id": 479489, "author_name": "karthik7395", "author_url": "", "post_date": "02/27/2019 07:31:20", "content": "You did really an amazing work @jingobox. You deserve a medal, always it is the best option to select two submissions with high CV and high LB.
", "votes": null, "replies": [] } ] }, { "id": 479485, "author_name": "karthik7395", "author_url": "", "post_date": "02/27/2019 07:28:43", "content": "Congratulations @petersk20 for getting your first medal.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "479070": "This is my first Kaggle competition so I want to know if the drastic changes between the public and private leader board are normal for a kaggle competition. I'm thinking that the change may be abnormal due to the number of outliers.", "479072": "There's a Shake up! Some top teams disappeared.", "479095": "That depends on the distribution of data in the train and test(hidden) data. \nif they have the same distribution, all goes well.\notherwise, there will be a big change between the public and private LB.\n\nThat's why Kaggle permit that you select two kernels for Your final score.\nSo select one that gives a good public score ie where you trust the public LB\nand another one that gives a good CV score because the CV score is more reliable if data changes.", "479107": "Yeah, there were too much shake ups. I wonder how some of the top teams at the public leaderboard managed to stay on top in the private leaderboard. Looking forward for their strategies that made robust submissions.", "479115": "It so happened that I didn't really trust CV, for which I was punished.", "479146": "So big gap with other teams, you make a amazing work.", "479155": "The outliers were really difficult to deal with, but, anyway, you did a great work!", "479485": "Congratulations @petersk20 for getting your first medal.", "479489": "You did really an amazing work @jingobox. You deserve a medal, always it is the best option to select two submissions with high CV and high LB." }, "source": "meta" }