Our proposal was: "To the admins: How can we take seriously the LB?"
\n\nWendy's answer was:
\n\n\n\n\nHi @George, @phunter
\n \nThank you for your suggestions. We will take some actions to mitigate.
\n \nEvery Kaggle competition is different. We do try our best to avoid\n data leakage, but it's not always perfectly avoidable given the\n natural constraints of the problem.
\n
It was 23 days ago.
\n\nWhere is the result of your "mitigation" or sorry if I missed something.\nThank you.
", "messages": [ { "id": "93333", "postDate": "09/23/2015 23:19:31", "content": "Our proposal was: "To the admins: How can we take seriously the LB?"
\n\nWendy's answer was:
\n\n\n\n\nHi @George, @phunter
\n \nThank you for your suggestions. We will take some actions to mitigate.
\n \nEvery Kaggle competition is different. We do try our best to avoid\n data leakage, but it's not always perfectly avoidable given the\n natural constraints of the problem.
\n
It was 23 days ago.
\n\nWhere is the result of your "mitigation" or sorry if I missed something.\nThank you.
", "rawMarkdown": "Our proposal was: \"To the admins: How can we take seriously the LB?\"\r\n\r\nWendy's answer was:\r\n\r\n> Hi @George, @phunter\r\n> \r\n> Thank you for your suggestions. We will take some actions to mitigate.\r\n> \r\n> Every Kaggle competition is different. We do try our best to avoid\r\n> data leakage, but it's not always perfectly avoidable given the\r\n> natural constraints of the problem.\r\n\r\nIt was 23 days ago.\r\n\r\nWhere is the result of your \"mitigation\" or sorry if I missed something.\r\nThank you.", "votes": null }, { "id": "93338", "postDate": "09/24/2015 00:14:04", "content": "@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
", "rawMarkdown": "George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.", "votes": null }, { "id": "93374", "postDate": "09/24/2015 16:57:19", "content": "[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
", "rawMarkdown": "[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.", "votes": null }, { "id": "93378", "postDate": "09/24/2015 17:45:18", "content": "[quote=George Solymosi;93374]
\n\n[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
\n\n[/quote]\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.
", "rawMarkdown": "[quote=George Solymosi;93374]\r\n\r\n[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.\r\n\r\n[/quote]\r\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.", "votes": null }, { "id": "93380", "postDate": "09/24/2015 18:12:02", "content": "[quote=phunter;93378]
\n\n[quote=George Solymosi;93374]
\n\n[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
\n\n[/quote]\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.
\n\n[/quote]
\n\nYou're absolutely right phunter and makes more sense of focusing to the main object.\nThough recall that it was 23 days ago, and we got no answer then at all.
", "rawMarkdown": "[quote=phunter;93378]\r\n\r\n[quote=George Solymosi;93374]\r\n\r\n[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.\r\n\r\n[/quote]\r\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.\r\n\r\n[/quote]\r\n\r\nYou're absolutely right phunter and makes more sense of focusing to the main object.\r\nThough recall that it was 23 days ago, and we got no answer then at all.", "votes": null }, { "id": "93383", "postDate": "09/24/2015 19:23:56", "content": "I see many down votes on George's post, and I am wondering what's happening.
\n\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.
\n\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman
", "rawMarkdown": "I see many down votes on George's post, and I am wondering what's happening. \r\n\r\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.\r\n\r\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman", "votes": null }, { "id": "93385", "postDate": "09/24/2015 20:48:16", "content": "[quote=phunter;93383]
\n\nI see many down votes on George's post, and I am wondering what's happening.
\n\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.
\n\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman
\n\n[/quote]
\n\n@phunter Thank you for your thoughts. I never was that kind of guy, who was ever afraid to tell his opinion or who was afraid to ask someone's help to integrate into my own knowledge fearing from other one's negative opinion (or downvotes). I always look forward and the time will tell unmercifully.
", "rawMarkdown": "[quote=phunter;93383]\r\n\r\nI see many down votes on George's post, and I am wondering what's happening. \r\n\r\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.\r\n\r\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman \r\n\r\n[/quote]\r\n\r\n@phunter Thank you for your thoughts. I never was that kind of guy, who was ever afraid to tell his opinion or who was afraid to ask someone's help to integrate into my own knowledge fearing from other one's negative opinion (or downvotes). I always look forward and the time will tell unmercifully.", "votes": null } ], "comments": [ { "id": 93338, "author_name": "wendykan", "author_url": "", "post_date": "09/24/2015 00:14:04", "content": "@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
", "votes": null, "replies": [] }, { "id": 93374, "author_name": "", "author_url": "", "post_date": "09/24/2015 16:57:19", "content": "[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
", "votes": null, "replies": [] }, { "id": 93378, "author_name": "phunter", "author_url": "", "post_date": "09/24/2015 17:45:18", "content": "[quote=George Solymosi;93374]
\n\n[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
\n\n[/quote]\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.
", "votes": null, "replies": [] }, { "id": 93380, "author_name": "", "author_url": "", "post_date": "09/24/2015 18:12:02", "content": "[quote=phunter;93378]
\n\n[quote=George Solymosi;93374]
\n\n[quote=Wendy Kan;93338]
\n\n@George,
\n\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our "mitigation". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.
\n\n[/quote]
\n\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.
\n\n[/quote]\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.
\n\n[/quote]
\n\nYou're absolutely right phunter and makes more sense of focusing to the main object.\nThough recall that it was 23 days ago, and we got no answer then at all.
", "votes": null, "replies": [] }, { "id": 93383, "author_name": "phunter", "author_url": "", "post_date": "09/24/2015 19:23:56", "content": "I see many down votes on George's post, and I am wondering what's happening.
\n\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.
\n\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman
", "votes": null, "replies": [] }, { "id": 93385, "author_name": "", "author_url": "", "post_date": "09/24/2015 20:48:16", "content": "[quote=phunter;93383]
\n\nI see many down votes on George's post, and I am wondering what's happening.
\n\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.
\n\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman
\n\n[/quote]
\n\n@phunter Thank you for your thoughts. I never was that kind of guy, who was ever afraid to tell his opinion or who was afraid to ask someone's help to integrate into my own knowledge fearing from other one's negative opinion (or downvotes). I always look forward and the time will tell unmercifully.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "93333": "Our proposal was: \"To the admins: How can we take seriously the LB?\"\r\n\r\nWendy's answer was:\r\n\r\n> Hi @George, @phunter\r\n> \r\n> Thank you for your suggestions. We will take some actions to mitigate.\r\n> \r\n> Every Kaggle competition is different. We do try our best to avoid\r\n> data leakage, but it's not always perfectly avoidable given the\r\n> natural constraints of the problem.\r\n\r\nIt was 23 days ago.\r\n\r\nWhere is the result of your \"mitigation\" or sorry if I missed something.\r\nThank you.", "93338": "George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before.", "93374": "[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.", "93378": "[quote=George Solymosi;93374]\r\n\r\n[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.\r\n\r\n[/quote]\r\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.", "93380": "[quote=phunter;93378]\r\n\r\n[quote=George Solymosi;93374]\r\n\r\n[quote=Wendy Kan;93338]\r\n\r\n@George,\r\n\r\nYes, we did make some inquiries. Responses were mixed and didn't come to a clear decision that help our \"mitigation\". After some internal discussions, we have decided to keep their submissions as they are now, since all kagglers deserve the right to compete as they wish, whether they will eventually win the competition or not. Also, since the top 2 on the leaderboard are no longer marked as out of the competition, we think it doesn't look as discouraging for new entrants as before. \r\n\r\n[/quote]\r\n\r\nThank you for your answer Wendy, though sorry to say, but it is actually not an answer at all and definitely not a correct mitigation.\r\n\r\n[/quote]\r\nI understand admins can have difficulties of removing competitors from the leaderboard before the deadline, since it may also discourage people. I just did my part of encouraging more people, by publishing my code skeleton for feature engineering plus grid search with xgboost. It can help many new comers getting into this competition. Xgboost has nice interface with scikit-learn and it has pip installation (install gcc firstly please), but no many people know it.\r\n\r\n[/quote]\r\n\r\nYou're absolutely right phunter and makes more sense of focusing to the main object.\r\nThough recall that it was 23 days ago, and we got no answer then at all.", "93383": "I see many down votes on George's post, and I am wondering what's happening. \r\n\r\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.\r\n\r\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman", "93385": "[quote=phunter;93383]\r\n\r\nI see many down votes on George's post, and I am wondering what's happening. \r\n\r\nCompetitions are for fun and learn, plus some reputation building. I believe most people here are not competing for the prize money: people who compete on Kaggle are great data scientists and they can easily make these money from their daily jobs. However, exploiting data leakage for winning is in the grey area: surely it would be a winning solution, but this winning solution would have no use for physics or to the Kaggle community (except a reminder for preventing future about data leakage). Having such winning solution may not help building individual's reputation in the data science community.\r\n\r\nExploiting the correlation as feedback to the training is not a hard secret, and I can post my experimental script using correlation anytime so everyone can jump to top 10. But I am a physicist (use to be, honestly, a difficulty story), and I believe “Physics is like sex: sure, it may give some practical results, but that's not why we do it.” ― Richard Feynman \r\n\r\n[/quote]\r\n\r\n@phunter Thank you for your thoughts. I never was that kind of guy, who was ever afraid to tell his opinion or who was afraid to ask someone's help to integrate into my own knowledge fearing from other one's negative opinion (or downvotes). I always look forward and the time will tell unmercifully." }, "source": "meta" }