It seems that the current test data set is included in the updated training dataset. So my submission achieves LB score0. Could you update the test data so that we can test how the updated models work?
", "messages": [ { "id": "1398480", "postDate": "07/24/2021 07:57:03", "content": "It seems that the current test data set is included in the updated training dataset. So my submission achieves LB score0. Could you update the test data so that we can test how the updated models work?
", "rawMarkdown": "It seems that the current test data set is included in the updated training dataset. So my submission achieves LB score0. Could you update the test data so that we can test how the updated models work?", "votes": null }, { "id": "1398902", "postDate": "07/24/2021 16:19:49", "content": "There is not enough data left for a new test dataset, as the newest data in train_updated comes from 2021-07-17, only a week ago.
", "rawMarkdown": "There is not enough data left for a new test dataset, as the newest data in train_updated comes from 2021-07-17, only a week ago.", "votes": null }, { "id": "1399065", "postDate": "07/24/2021 20:49:42", "content": "@juliaelliott Hi, could you please delete our latest submission which scored 0 on the LB because of the data leak.
", "rawMarkdown": "juliaelliott Hi, could you please delete our latest submission which scored 0 on the LB because of the data leak.", "votes": null }, { "id": "1400923", "postDate": "07/26/2021 17:57:45", "content": "@huiyujiang95 No, we do not delete submissions. Note for you and other participants: Any advancement you make on the public leaderboard using the updated training data that was provided (whose ground truth overlapped with the current public leaderboard) will not have any improvement or impact on your final standing, as the evaluation period which determines your final ranking will be a future timeframe, for which the ground truth is never provided.
", "rawMarkdown": "huiyujiang95 No, we do not delete submissions. Note for you and other participants: Any advancement you make on the public leaderboard using the updated training data that was provided (whose ground truth overlapped with the current public leaderboard) will not have any improvement or impact on your final standing, as the evaluation period which determines your final ranking will be a future timeframe, for which the ground truth is never provided.", "votes": null }, { "id": "1400928", "postDate": "07/26/2021 18:01:42", "content": "Why downvoting this topic? \"Be patient. Be friendly. We are all here to learn. \".
\nThere is no reason to submit predictions now unless you want to check your mae score estimate is the same as kaggle platform PL computation. For this, use models trained on data before May to predict May targets. The score on such predictions should be equal to PL score.
\nTo train better models on new (incl. July data) just rely on your validation score and please Do Not Submit.
", "rawMarkdown": "Why downvoting this topic? \"Be patient. Be friendly. We are all here to learn. \".\n\nThere is no reason to submit predictions now unless you want to check your mae score estimate is the same as kaggle platform PL computation. For this, use models trained on data before May to predict May targets. The score on such predictions should be equal to PL score. \n\nTo train better models on new (incl. July data) just rely on your validation score and please Do Not Submit.", "votes": null } ], "comments": [ { "id": 1398902, "author_name": "jiangtt", "author_url": "", "post_date": "07/24/2021 16:19:49", "content": "There is not enough data left for a new test dataset, as the newest data in train_updated comes from 2021-07-17, only a week ago.
", "votes": null, "replies": [] }, { "id": 1399065, "author_name": "huiyujiang95", "author_url": "", "post_date": "07/24/2021 20:49:42", "content": "@juliaelliott Hi, could you please delete our latest submission which scored 0 on the LB because of the data leak.
", "votes": null, "replies": [ { "id": 1400923, "author_name": "juliaelliott", "author_url": "", "post_date": "07/26/2021 17:57:45", "content": "@huiyujiang95 No, we do not delete submissions. Note for you and other participants: Any advancement you make on the public leaderboard using the updated training data that was provided (whose ground truth overlapped with the current public leaderboard) will not have any improvement or impact on your final standing, as the evaluation period which determines your final ranking will be a future timeframe, for which the ground truth is never provided.
", "votes": null, "replies": [] } ] }, { "id": 1400928, "author_name": "isakev", "author_url": "", "post_date": "07/26/2021 18:01:42", "content": "Why downvoting this topic? \"Be patient. Be friendly. We are all here to learn. \".
\nThere is no reason to submit predictions now unless you want to check your mae score estimate is the same as kaggle platform PL computation. For this, use models trained on data before May to predict May targets. The score on such predictions should be equal to PL score.
\nTo train better models on new (incl. July data) just rely on your validation score and please Do Not Submit.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1398480": "It seems that the current test data set is included in the updated training dataset. So my submission achieves LB score0. Could you update the test data so that we can test how the updated models work?", "1398902": "There is not enough data left for a new test dataset, as the newest data in train_updated comes from 2021-07-17, only a week ago.", "1399065": "juliaelliott Hi, could you please delete our latest submission which scored 0 on the LB because of the data leak.", "1400923": "huiyujiang95 No, we do not delete submissions. Note for you and other participants: Any advancement you make on the public leaderboard using the updated training data that was provided (whose ground truth overlapped with the current public leaderboard) will not have any improvement or impact on your final standing, as the evaluation period which determines your final ranking will be a future timeframe, for which the ground truth is never provided.", "1400928": "Why downvoting this topic? \"Be patient. Be friendly. We are all here to learn. \".\n\nThere is no reason to submit predictions now unless you want to check your mae score estimate is the same as kaggle platform PL computation. For this, use models trained on data before May to predict May targets. The score on such predictions should be equal to PL score. \n\nTo train better models on new (incl. July data) just rely on your validation score and please Do Not Submit." }, "source": "meta" }