I am really wondering if it's a good idea for kaggle to release private test set after the competition ends. In my opinion it will help in gaining better understanding of what we did wrong and can help in improving our pipelines for future competitions.
\n\nIf you guys are going to downvote this please let me know what's wrong in this discussion post.
", "messages": [ { "id": 941976, "postDate": "2020-07-23T14:23:20.683Z", "content": "I am really wondering if it's a good idea for kaggle to release private test set after the competition ends. In my opinion it will help in gaining better understanding of what we did wrong and can help in improving our pipelines for future competitions.
\n\nIf you guys are going to downvote this please let me know what's wrong in this discussion post.
", "rawMarkdown": "I am really wondering if it's a good idea for kaggle to release private test set after the competition ends. In my opinion it will help in gaining better understanding of what we did wrong and can help in improving our pipelines for future competitions.\n\nIf you guys are going to downvote this please let me know what's wrong in this discussion post.", "votes": 12 }, { "id": 942150, "postDate": "2020-07-23T15:58:55.450Z", "content": "It should be kept private.
", "rawMarkdown": "It should be kept private.", "votes": 3, "replies": [ { "id": 942414, "postDate": "2020-07-23T18:21:13.287Z", "content": "Would you elaborate in your answer?
", "rawMarkdown": "Would you elaborate in your answer?" }, { "id": 942625, "postDate": "2020-07-23T22:08:45.477Z", "content": "The whole point was to see if you can train models on some data then use it on data labelled by other radiologists. As soon as you expose test data you no longer can test this. It means that this dataset that took years to assemble becomes almost useless.
\n\nThink of this as imagenet for prostate cancer. Let's make it useful for several years rather then a one time thing.
", "rawMarkdown": "The whole point was to see if you can train models on some data then use it on data labelled by other radiologists. As soon as you expose test data you no longer can test this. It means that this dataset that took years to assemble becomes almost useless.\n\nThink of this as imagenet for prostate cancer. Let's make it useful for several years rather then a one time thing.", "votes": 7 } ] }, { "id": 945131, "postDate": "2020-07-25T15:33:43.087Z", "content": "I don't know. Does that really matter? What difference does that bring? I think we can deduce enough insights from the available data itself.
", "rawMarkdown": "I don't know. Does that really matter? What difference does that bring? I think we can deduce enough insights from the available data itself. " }, { "id": 942065, "postDate": "2020-07-23T15:04:33.667Z", "content": "You are so right, but, it depends more on the host of the competition than Kaggle by itself isn't?
", "rawMarkdown": "You are so right, but, it depends more on the host of the competition than Kaggle by itself isn't?", "replies": [ { "id": 942119, "postDate": "2020-07-23T15:36:34.043Z", "content": "@hiramcho yes, you are right. It's more of a host's call then Kaggle's.
", "rawMarkdown": "@hiramcho yes, you are right. It's more of a host's call then Kaggle's." } ] } ], "comments": [ { "id": 942150, "author_name": "CPMP", "author_url": "", "post_date": "2020-07-23T15:58:55.450000", "content": "It should be kept private.
", "votes": 3, "replies": [ { "id": 942414, "author_name": "Hiram Coria 🧬", "author_url": "", "post_date": "2020-07-23T18:21:13.287000", "content": "Would you elaborate in your answer?
", "votes": 0, "replies": [] }, { "id": 942625, "author_name": "CPMP", "author_url": "", "post_date": "2020-07-23T22:08:45.477000", "content": "The whole point was to see if you can train models on some data then use it on data labelled by other radiologists. As soon as you expose test data you no longer can test this. It means that this dataset that took years to assemble becomes almost useless.
\n\nThink of this as imagenet for prostate cancer. Let's make it useful for several years rather then a one time thing.
", "votes": 7, "replies": [] } ] }, { "id": 945131, "author_name": "Jaseem C K", "author_url": "", "post_date": "2020-07-25T15:33:43.087000", "content": "I don't know. Does that really matter? What difference does that bring? I think we can deduce enough insights from the available data itself.
", "votes": 0, "replies": [] }, { "id": 942065, "author_name": "Hiram Coria 🧬", "author_url": "", "post_date": "2020-07-23T15:04:33.667000", "content": "You are so right, but, it depends more on the host of the competition than Kaggle by itself isn't?
", "votes": 0, "replies": [ { "id": 942119, "author_name": "Dracarys", "author_url": "", "post_date": "2020-07-23T15:36:34.043000", "content": "@hiramcho yes, you are right. It's more of a host's call then Kaggle's.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "941976": "I am really wondering if it's a good idea for kaggle to release private test set after the competition ends. In my opinion it will help in gaining better understanding of what we did wrong and can help in improving our pipelines for future competitions.\n\nIf you guys are going to downvote this please let me know what's wrong in this discussion post.", "942150": "It should be kept private.", "945131": "I don't know. Does that really matter? What difference does that bring? I think we can deduce enough insights from the available data itself. ", "942065": "You are so right, but, it depends more on the host of the competition than Kaggle by itself isn't?" } }