Did you guys see the top1 on public lb?
\nscore 0.6, seriously?
\nCan you show some MAGIC?
", "messages": [ { "id": 1398095, "postDate": "2021-07-23T18:09:50.340Z", "content": "Now that the answers for May have been provided via the train_updated.csv file the Public LB doesn't mean anything now. Someone could probably post a 0.0 if they wanted. Only LB that matters is the live LB in August/September
It is quite weird for train_updated.csv to have the data of May.
\nThen the public LB has no reference value ANY MORE.
This means: WE CANNOT see the relative rank right now, but we can only guess our ranking through our intuition.
", "rawMarkdown": "It is quite weird for train_updated.csv to have the data of May.\nThen the public LB has no reference value ANY MORE. \n\nThis means: WE CANNOT see the relative rank right now, but we can only guess our ranking through our intuition." }, { "id": 1398263, "postDate": "2021-07-24T00:08:58.567Z", "content": "I understand that some people may favor other factors, but I'm happy with this. We have all the data available as opposed to having some of it hidden and just being able to use it 5 times a day (submissions), without the ability to process it as we see fit (e.g. use just part of it or use it in combination with other data).
\nWe do lose the LB, but I personally never cared much about where I stand in it. I suppose we all have a feeling of how much our models might be overfitting, but we really have no idea about the others. We just have to do the work and be patient. In time, we'll know where we stand :).
", "rawMarkdown": "I understand that some people may favor other factors, but I'm happy with this. We have all the data available as opposed to having some of it hidden and just being able to use it 5 times a day (submissions), without the ability to process it as we see fit (e.g. use just part of it or use it in combination with other data).\n\nWe do lose the LB, but I personally never cared much about where I stand in it. I suppose we all have a feeling of how much our models might be overfitting, but we really have no idea about the others. We just have to do the work and be patient. In time, we'll know where we stand :).", "votes": 2 } ] }, { "id": 1398092, "postDate": "2021-07-23T18:07:22.733Z", "content": "Did you guys see the top1 on public lb?
\nscore 0.6, seriously?
\nCan you show some MAGIC?
", "rawMarkdown": "Did you guys see the top1 on public lb?\n\nscore 0.6, seriously? \n\nCan you show some MAGIC?" }, { "id": 1398543, "postDate": "2021-07-24T09:12:50.240Z", "content": "there is score 0.0 im suprized. i think there is a leak.
", "rawMarkdown": "there is score 0.0 im suprized. i think there is a leak.", "votes": 1, "replies": [ { "id": 1398555, "postDate": "2021-07-24T09:25:46.627Z", "content": "++++++++++..
", "rawMarkdown": "++++++++++.." }, { "id": 1398578, "postDate": "2021-07-24T09:46:30.657Z", "content": "Ground truth of may data was released in the updated training.
\nI was surprised why it took more than a day till someone submitted perfect score.
In SETI comp before reseting, it took less than a day.
\nAs someone already stated, the lb ranking is meaningless now.
", "rawMarkdown": "Ground truth of may data was released in the updated training.\nI was surprised why it took more than a day till someone submitted perfect score. \n\nIn SETI comp before reseting, it took less than a day.\n\nAs someone already stated, the lb ranking is meaningless now. ", "votes": 1 }, { "id": 1398591, "postDate": "2021-07-24T10:03:30.090Z", "content": "everything will be visible on pr lb
", "rawMarkdown": "everything will be visible on pr lb", "votes": 1 }, { "id": 1399450, "postDate": "2021-07-25T10:08:45.713Z", "content": "We have to concentrate competition. 😄
", "rawMarkdown": "We have to concentrate competition. 😄", "votes": 1 } ] }, { "id": 1398094, "postDate": "2021-07-23T18:09:00.993Z", "content": "This is incredibly unbelievable.
\nWhat is much more amazing is that: he only submitted one entry..👀
If so do for good ranking, it mean we need to use training targets to test target in trained date range, then apply model after out of the trained date range?
\nor final ranking will base on the result other than trained data?
p.s. this is my first competition, so I do not know the detail on final Leaderboard.
", "rawMarkdown": "If so do for good ranking, it mean we need to use training targets to test target in trained date range, then apply model after out of the trained date range?\nor final ranking will base on the result other than trained data?\n\np.s. this is my first competition, so I do not know the detail on final Leaderboard.", "replies": [ { "id": 1399083, "postDate": "2021-07-24T21:45:51.930Z", "content": "In this competition getting better score on the public leaderboard doesn't imply the better scores on the private leaderboard since the evaluate time interval are different.
\nThe final ranking result is based on the targets after 31th July.
thanks for clearing on that!
", "rawMarkdown": "thanks for clearing on that!" } ] }, { "id": 1398348, "postDate": "2021-07-24T05:00:14.253Z", "content": "btw, as the external data set need to be Freely & publicly available, is it available to see other dataset used in submission?
", "rawMarkdown": "btw, as the external data set need to be Freely & publicly available, is it available to see other dataset used in submission?", "replies": [ { "id": 1398364, "postDate": "2021-07-24T05:24:06.343Z", "content": "Probably not
", "rawMarkdown": "Probably not" } ] }, { "id": 1398272, "postDate": "2021-07-24T00:39:52.720Z", "content": "As a somebody who has decided to come back into the competition during the last week, what is the valid way to validate my results now?
\nShould I manually drop newly added dates and consider such result to be valid?
\nGuess that I should add a \"valid\" word few more times, eww.
", "rawMarkdown": "As a somebody who has decided to come back into the competition during the last week, what is the valid way to validate my results now? \n\nShould I manually drop newly added dates and consider such result to be valid?\n\nGuess that I should add a \"valid\" word few more times, eww." }, { "id": 1399449, "postDate": "2021-07-25T10:05:13.093Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1399224, "postDate": "2021-07-25T03:45:11.530Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1398095, "author_name": "Branden Murray", "author_url": "", "post_date": "2021-07-23T18:09:50.340000", "content": "Now that the answers for May have been provided via the train_updated.csv file the Public LB doesn't mean anything now. Someone could probably post a 0.0 if they wanted. Only LB that matters is the live LB in August/September
It is quite weird for train_updated.csv to have the data of May.
\nThen the public LB has no reference value ANY MORE.
This means: WE CANNOT see the relative rank right now, but we can only guess our ranking through our intuition.
", "votes": 0, "replies": [] }, { "id": 1398263, "author_name": "vialactea", "author_url": "", "post_date": "2021-07-24T00:08:58.567000", "content": "I understand that some people may favor other factors, but I'm happy with this. We have all the data available as opposed to having some of it hidden and just being able to use it 5 times a day (submissions), without the ability to process it as we see fit (e.g. use just part of it or use it in combination with other data).
\nWe do lose the LB, but I personally never cared much about where I stand in it. I suppose we all have a feeling of how much our models might be overfitting, but we really have no idea about the others. We just have to do the work and be patient. In time, we'll know where we stand :).
", "votes": 2, "replies": [] } ] }, { "id": 1398543, "author_name": "tensor choko", "author_url": "", "post_date": "2021-07-24T09:12:50.240000", "content": "there is score 0.0 im suprized. i think there is a leak.
", "votes": 1, "replies": [ { "id": 1398555, "author_name": "Radmir Zosimov", "author_url": "", "post_date": "2021-07-24T09:25:46.627000", "content": "++++++++++..
", "votes": 0, "replies": [] }, { "id": 1398578, "author_name": "SHih Chieh Lai", "author_url": "", "post_date": "2021-07-24T09:46:30.657000", "content": "Ground truth of may data was released in the updated training.
\nI was surprised why it took more than a day till someone submitted perfect score.
In SETI comp before reseting, it took less than a day.
\nAs someone already stated, the lb ranking is meaningless now.
", "votes": 1, "replies": [] }, { "id": 1398591, "author_name": "Radmir Zosimov", "author_url": "", "post_date": "2021-07-24T10:03:30.090000", "content": "everything will be visible on pr lb
", "votes": 1, "replies": [] }, { "id": 1399450, "author_name": "tensor choko", "author_url": "", "post_date": "2021-07-25T10:08:45.713000", "content": "We have to concentrate competition. 😄
", "votes": 1, "replies": [] } ] }, { "id": 1398094, "author_name": "Elliot", "author_url": "", "post_date": "2021-07-23T18:09:00.993000", "content": "This is incredibly unbelievable.
\nWhat is much more amazing is that: he only submitted one entry..👀
If so do for good ranking, it mean we need to use training targets to test target in trained date range, then apply model after out of the trained date range?
\nor final ranking will base on the result other than trained data?
p.s. this is my first competition, so I do not know the detail on final Leaderboard.
", "votes": 0, "replies": [ { "id": 1399083, "author_name": "JongOS", "author_url": "", "post_date": "2021-07-24T21:45:51.930000", "content": "In this competition getting better score on the public leaderboard doesn't imply the better scores on the private leaderboard since the evaluate time interval are different.
\nThe final ranking result is based on the targets after 31th July.
thanks for clearing on that!
", "votes": 0, "replies": [] } ] }, { "id": 1398348, "author_name": "P. L", "author_url": "", "post_date": "2021-07-24T05:00:14.253000", "content": "btw, as the external data set need to be Freely & publicly available, is it available to see other dataset used in submission?
", "votes": 0, "replies": [ { "id": 1398364, "author_name": "Zacchaeus", "author_url": "", "post_date": "2021-07-24T05:24:06.343000", "content": "Probably not
", "votes": 0, "replies": [] } ] }, { "id": 1398272, "author_name": "majoraregalia", "author_url": "", "post_date": "2021-07-24T00:39:52.720000", "content": "As a somebody who has decided to come back into the competition during the last week, what is the valid way to validate my results now?
\nShould I manually drop newly added dates and consider such result to be valid?
\nGuess that I should add a \"valid\" word few more times, eww.
", "votes": 0, "replies": [] }, { "id": 1399449, "author_name": "", "author_url": "", "post_date": "2021-07-25T10:05:13.093000", "content": "", "votes": 0, "replies": [] }, { "id": 1399224, "author_name": "", "author_url": "", "post_date": "2021-07-25T03:45:11.530000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1398095": "Now that the answers for May have been provided via the `train_updated.csv` file the Public LB doesn't mean anything now. Someone could probably post a 0.0 if they wanted. Only LB that matters is the live LB in August/September", "1398092": "Did you guys see the top1 on public lb?\n\nscore 0.6, seriously? \n\nCan you show some MAGIC?", "1398543": "there is score 0.0 im suprized. i think there is a leak.", "1398094": "This is incredibly unbelievable.\nWhat is much more amazing is that: he only submitted one entry..👀", "1398602": "If so do for good ranking, it mean we need to use training targets to test target in trained date range, then apply model after out of the trained date range?\nor final ranking will base on the result other than trained data?\n\np.s. this is my first competition, so I do not know the detail on final Leaderboard.", "1398348": "btw, as the external data set need to be Freely & publicly available, is it available to see other dataset used in submission?", "1398272": "As a somebody who has decided to come back into the competition during the last week, what is the valid way to validate my results now? \n\nShould I manually drop newly added dates and consider such result to be valid?\n\nGuess that I should add a \"valid\" word few more times, eww.", "1399449": "", "1399224": "" } }