If KAGGLE don't do something to prevent publishing high score kernerls before last days, i will NOT spend time on kaggle competitons anymore and prevent other friends to pay effort on this. 42 hours GPU limits is defenitely not enough to train a new model. It's unfair for our time and effort on it. I wish kaggle staff cound reply on this.
", "messages": [ { "id": 1058915, "postDate": "2020-10-24T12:16:37.387Z", "content": "If KAGGLE don't do something to prevent publishing high score kernerls before last days, i will NOT spend time on kaggle competitons anymore and prevent other friends to pay effort on this. 42 hours GPU limits is defenitely not enough to train a new model. It's unfair for our time and effort on it. I wish kaggle staff cound reply on this.
", "rawMarkdown": "If KAGGLE don't do something to prevent publishing high score kernerls before last days, i will NOT spend time on kaggle competitons anymore and prevent other friends to pay effort on this. 42 hours GPU limits is defenitely not enough to train a new model. It's unfair for our time and effort on it. I wish kaggle staff cound reply on this.", "votes": 16 }, { "id": 1058963, "postDate": "2020-10-24T13:05:59.490Z", "content": "Now the whole competition is a COMPETE of COMPUTING resources. With the new public kernel, those who has strong computing resources could just get a medal with \"copy and train\". I don't think it's the meaning of kaggle competition.
", "rawMarkdown": "Now the whole competition is a COMPETE of COMPUTING resources. With the new public kernel, those who has strong computing resources could just get a medal with \"copy and train\". I don't think it's the meaning of kaggle competition.", "votes": 10, "replies": [ { "id": 1059932, "postDate": "2020-10-25T15:46:20.447Z", "content": "And in the same time, people who just forked a high score notebook will get medals -- and without any merit and effort, as they compete with somebody's else's work. When one sees 100 or 200 competitors with exactly the same score, which is also the score of a high scored public notebook, THAT is suspicious. Moreover, most of these people who most probably forked the high scored notebook, are novices and they have only one or two entries.
\nInspiring oneself from a good notebook, taking some ideas and applying them to one's own model is a thing. But making a copy/paste or a fork and competing with someone's else's work only, that makes no sense and is an unfair disadvantage for other teams who worked hard.
", "rawMarkdown": "And in the same time, people who just forked a high score notebook will get medals -- and without any merit and effort, as they compete with somebody's else's work. When one sees 100 or 200 competitors with exactly the same score, which is also the score of a high scored public notebook, THAT is suspicious. Moreover, most of these people who most probably forked the high scored notebook, are novices and they have only one or two entries. \n\nInspiring oneself from a good notebook, taking some ideas and applying them to one's own model is a thing. But making a copy/paste or a fork and competing with someone's else's work only, that makes no sense and is an unfair disadvantage for other teams who worked hard.", "votes": 4 } ] }, { "id": 1058960, "postDate": "2020-10-24T13:00:07.540Z", "content": "I think they should change their prohibition policy of publishing high scoring kernels from 1 week to atleast 2 or 3 weeks before competition deadline. And should take strict actions to enforce that.
", "rawMarkdown": "I think they should change their prohibition policy of publishing high scoring kernels from 1 week to atleast 2 or 3 weeks before competition deadline. And should take strict actions to enforce that. ", "votes": 8, "replies": [ { "id": 1059077, "postDate": "2020-10-24T15:25:50.690Z", "content": "Yeah, you are absolutely right, kaggle must change some rules about high scoring kernel for last 1 or 2 weeks before the deadline.
", "rawMarkdown": "Yeah, you are absolutely right, kaggle must change some rules about high scoring kernel for last 1 or 2 weeks before the deadline.", "votes": 4 }, { "id": 1061245, "postDate": "2020-10-26T21:13:09.673Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 1059272, "postDate": "2020-10-24T21:02:04.370Z", "content": "such a shame. it was going to be my first competition. And I was going to implement a similar idea too! All I have now is an idea that I pray that can at least give me a bronze of a silver.
", "rawMarkdown": "such a shame. it was going to be my first competition. And I was going to implement a similar idea too! All I have now is an idea that I pray that can at least give me a bronze of a silver.", "votes": -12 } ], "comments": [ { "id": 1058963, "author_name": "Gzl0506", "author_url": "", "post_date": "2020-10-24T13:05:59.490000", "content": "Now the whole competition is a COMPETE of COMPUTING resources. With the new public kernel, those who has strong computing resources could just get a medal with \"copy and train\". I don't think it's the meaning of kaggle competition.
", "votes": 10, "replies": [ { "id": 1059932, "author_name": "Catadanna", "author_url": "", "post_date": "2020-10-25T15:46:20.447000", "content": "And in the same time, people who just forked a high score notebook will get medals -- and without any merit and effort, as they compete with somebody's else's work. When one sees 100 or 200 competitors with exactly the same score, which is also the score of a high scored public notebook, THAT is suspicious. Moreover, most of these people who most probably forked the high scored notebook, are novices and they have only one or two entries.
\nInspiring oneself from a good notebook, taking some ideas and applying them to one's own model is a thing. But making a copy/paste or a fork and competing with someone's else's work only, that makes no sense and is an unfair disadvantage for other teams who worked hard.
", "votes": 4, "replies": [] } ] }, { "id": 1058960, "author_name": "Dracarys", "author_url": "", "post_date": "2020-10-24T13:00:07.540000", "content": "I think they should change their prohibition policy of publishing high scoring kernels from 1 week to atleast 2 or 3 weeks before competition deadline. And should take strict actions to enforce that.
", "votes": 8, "replies": [ { "id": 1059077, "author_name": "", "author_url": "", "post_date": "2020-10-24T15:25:50.690000", "content": "Yeah, you are absolutely right, kaggle must change some rules about high scoring kernel for last 1 or 2 weeks before the deadline.
", "votes": 4, "replies": [] }, { "id": 1061245, "author_name": "", "author_url": "", "post_date": "2020-10-26T21:13:09.673000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1059272, "author_name": "DarkCube", "author_url": "", "post_date": "2020-10-24T21:02:04.370000", "content": "such a shame. it was going to be my first competition. And I was going to implement a similar idea too! All I have now is an idea that I pray that can at least give me a bronze of a silver.
", "votes": -12, "replies": [] } ], "raw_markdown_by_id": { "1058915": "If KAGGLE don't do something to prevent publishing high score kernerls before last days, i will NOT spend time on kaggle competitons anymore and prevent other friends to pay effort on this. 42 hours GPU limits is defenitely not enough to train a new model. It's unfair for our time and effort on it. I wish kaggle staff cound reply on this.", "1058963": "Now the whole competition is a COMPETE of COMPUTING resources. With the new public kernel, those who has strong computing resources could just get a medal with \"copy and train\". I don't think it's the meaning of kaggle competition.", "1058960": "I think they should change their prohibition policy of publishing high scoring kernels from 1 week to atleast 2 or 3 weeks before competition deadline. And should take strict actions to enforce that. ", "1059272": "such a shame. it was going to be my first competition. And I was going to implement a similar idea too! All I have now is an idea that I pray that can at least give me a bronze of a silver." } }