![\"\"](\"https://i.postimg.cc/R0DySXXK/Screenshot-20220510-113106-Chrome.jpg\")
We can easliy find csci567_id accounts in leaderboard.
\n
We can easliy find csci567_id accounts in leaderboard.
\n
If you suspect cheating, you can send a mail to compliance@kaggle.com.
", "rawMarkdown": "If you suspect cheating, you can send a mail to compliance@kaggle.com.", "votes": null }, { "id": "1783181", "postDate": "05/10/2022 07:17:34", "content": "That is because our class is using this competition as our final project.
", "rawMarkdown": "That is because our class is using this competition as our final project.", "votes": null }, { "id": "1783418", "postDate": "05/10/2022 11:43:28", "content": "\nDo you mean your class in a paid tutoring service similar to that described by ZZY and others in https://www.kaggle.com/competitions/happy-whale-and-dolphin/discussion/319650 ?
\nBTW your team is a perfect example for my soon coming statistics, because you are a classic type of \"instant genius\" - from nowhere suddenly to silver with only 11 submissions while your teammates are classic types of \"silent companions\" - earning silver with zero respectively 1 day of activity on Kaggle.
", "rawMarkdown": "edddddd \n\nDo you mean your class in a paid tutoring service similar to that described by ZZY and others in [https://www.kaggle.com/competitions/happy-whale-and-dolphin/discussion/319650 ](url) ?\n\nBTW your team is a perfect example for my soon coming statistics, because you are a classic type of \"instant genius\" - from nowhere suddenly to silver with only 11 submissions while your teammates are classic types of \"silent companions\" - earning silver with zero respectively 1 day of activity on Kaggle.", "votes": null }, { "id": "1783667", "postDate": "05/10/2022 15:09:01", "content": "It seems like finding cheaters might be great next competition from Kaggle 🤥
", "rawMarkdown": "It seems like finding cheaters might be great next competition from Kaggle 🤥", "votes": null }, { "id": "1783680", "postDate": "05/10/2022 15:22:02", "content": "Since this is a project and also a competition, there is no code sharing at least from what I know. Our class is about the theoretical learning of machine learning, I guess many students are acutally struggling with this project.
\nAnd I was trying to build a ranking model described in the discusstion 307288. It has tremendous useful informations about how we should build ranking models. I originally did some experiments using some static features on one week transaction, but the cv scores were not better than the method from public scores. This is my first time working on the ranking model, so I am not confident about if I am on the right direction. So I spent lots of time researching about the ranking model implementations. There were many details that I wasn't sure. Our team was about to use the ensemble method described here as our final submission. And there was still one day left, so I decided to improve my ranking model based on my understanding, and see how it goes.
\nAfter I added more features like the counts, popularity, and days, my ranking model improved. Then I added more recall methods from public notebooks (since I don't have enough time to build one myself) to generate more condidates, and train them on last 6 week instead of 1 week. The local cv improved a lot. Then I spent some time adding more simple count-based featuers. I can finish them in one day because I was actually in the right direction previously, and the rest of work was built on top of my previous version.
\nSo I think I am just lucky to get the scores beyond my expectation.
", "rawMarkdown": "Since this is a project and also a competition, there is no code sharing at least from what I know. Our class is about the theoretical learning of machine learning, I guess many students are acutally struggling with this project.\n\nAnd I was trying to build a ranking model described in the discusstion [307288](https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/307288). It has tremendous useful informations about how we should build ranking models. I originally did some experiments using some static features on one week transaction, but the cv scores were not better than the method from public scores. This is my first time working on the ranking model, so I am not confident about if I am on the right direction. So I spent lots of time researching about the ranking model implementations. There were many details that I wasn't sure. Our team was about to use the ensemble method described [here](https://www.kaggle.com/code/tarique7/lb-0-0240-h-m-ensemble-magic-multi-blend) as our final submission. And there was still one day left, so I decided to improve my ranking model based on my understanding, and see how it goes. \n\nAfter I added more features like the counts, popularity, and days, my ranking model improved. Then I added more recall methods from public notebooks (since I don't have enough time to build one myself) to generate more condidates, and train them on last 6 week instead of 1 week. The local cv improved a lot. Then I spent some time adding more simple count-based featuers. I can finish them in one day because I was actually in the right direction previously, and the rest of work was built on top of my previous version. \n\nSo I think I am just lucky to get the scores beyond my expectation.", "votes": null }, { "id": "1784635", "postDate": "05/11/2022 10:38:32", "content": "@edddddd
\nAs a write up it is just chaotic but the crucial problem is that you didn't answer my questions.
First - which school or institution is your \"class\" from?
\nSecond - your official teammates should get a silver medal for 1 respectively zero days of activity on Kaggle. Is this your idea of fairness in Kaggle competitions?
", "rawMarkdown": "edddddd \nAs a write up it is just chaotic but the crucial problem is that you didn't answer my questions.\n\nFirst - **which school** or institution is your \"class\" from?\n\nSecond - your **official teammates** should get a silver medal **for 1 respectively zero days** of activity on Kaggle. Is this your idea of fairness in Kaggle competitions?", "votes": null }, { "id": "1784761", "postDate": "05/11/2022 12:53:55", "content": "many of them even dont have any submission
", "rawMarkdown": "many of them even dont have any submission", "votes": null } ], "comments": [ { "id": 1782957, "author_name": "zhoumichael", "author_url": "", "post_date": "05/10/2022 03:00:28", "content": "If you suspect cheating, you can send a mail to compliance@kaggle.com.
", "votes": null, "replies": [] }, { "id": 1783181, "author_name": "edddddd", "author_url": "", "post_date": "05/10/2022 07:17:34", "content": "That is because our class is using this competition as our final project.
", "votes": null, "replies": [ { "id": 1783418, "author_name": "blankaf", "author_url": "", "post_date": "05/10/2022 11:43:28", "content": "\nDo you mean your class in a paid tutoring service similar to that described by ZZY and others in https://www.kaggle.com/competitions/happy-whale-and-dolphin/discussion/319650 ?
\nBTW your team is a perfect example for my soon coming statistics, because you are a classic type of \"instant genius\" - from nowhere suddenly to silver with only 11 submissions while your teammates are classic types of \"silent companions\" - earning silver with zero respectively 1 day of activity on Kaggle.
", "votes": null, "replies": [] }, { "id": 1783680, "author_name": "edddddd", "author_url": "", "post_date": "05/10/2022 15:22:02", "content": "Since this is a project and also a competition, there is no code sharing at least from what I know. Our class is about the theoretical learning of machine learning, I guess many students are acutally struggling with this project.
\nAnd I was trying to build a ranking model described in the discusstion 307288. It has tremendous useful informations about how we should build ranking models. I originally did some experiments using some static features on one week transaction, but the cv scores were not better than the method from public scores. This is my first time working on the ranking model, so I am not confident about if I am on the right direction. So I spent lots of time researching about the ranking model implementations. There were many details that I wasn't sure. Our team was about to use the ensemble method described here as our final submission. And there was still one day left, so I decided to improve my ranking model based on my understanding, and see how it goes.
\nAfter I added more features like the counts, popularity, and days, my ranking model improved. Then I added more recall methods from public notebooks (since I don't have enough time to build one myself) to generate more condidates, and train them on last 6 week instead of 1 week. The local cv improved a lot. Then I spent some time adding more simple count-based featuers. I can finish them in one day because I was actually in the right direction previously, and the rest of work was built on top of my previous version.
\nSo I think I am just lucky to get the scores beyond my expectation.
", "votes": null, "replies": [] }, { "id": 1784635, "author_name": "blankaf", "author_url": "", "post_date": "05/11/2022 10:38:32", "content": "@edddddd
\nAs a write up it is just chaotic but the crucial problem is that you didn't answer my questions.
First - which school or institution is your \"class\" from?
\nSecond - your official teammates should get a silver medal for 1 respectively zero days of activity on Kaggle. Is this your idea of fairness in Kaggle competitions?
", "votes": null, "replies": [] } ] }, { "id": 1783667, "author_name": "lukawskikacper", "author_url": "", "post_date": "05/10/2022 15:09:01", "content": "It seems like finding cheaters might be great next competition from Kaggle 🤥
", "votes": null, "replies": [] }, { "id": 1784761, "author_name": "poverlu", "author_url": "", "post_date": "05/11/2022 12:53:55", "content": "many of them even dont have any submission
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1782948": "We can easliy find csci567_id accounts in leaderboard. \n\n", "1782957": "If you suspect cheating, you can send a mail to compliance@kaggle.com.", "1783181": "That is because our class is using this competition as our final project.", "1783418": "edddddd \n\nDo you mean your class in a paid tutoring service similar to that described by ZZY and others in [https://www.kaggle.com/competitions/happy-whale-and-dolphin/discussion/319650 ](url) ?\n\nBTW your team is a perfect example for my soon coming statistics, because you are a classic type of \"instant genius\" - from nowhere suddenly to silver with only 11 submissions while your teammates are classic types of \"silent companions\" - earning silver with zero respectively 1 day of activity on Kaggle.", "1783667": "It seems like finding cheaters might be great next competition from Kaggle 🤥", "1783680": "Since this is a project and also a competition, there is no code sharing at least from what I know. Our class is about the theoretical learning of machine learning, I guess many students are acutally struggling with this project.\n\nAnd I was trying to build a ranking model described in the discusstion [307288](https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/307288). It has tremendous useful informations about how we should build ranking models. I originally did some experiments using some static features on one week transaction, but the cv scores were not better than the method from public scores. This is my first time working on the ranking model, so I am not confident about if I am on the right direction. So I spent lots of time researching about the ranking model implementations. There were many details that I wasn't sure. Our team was about to use the ensemble method described [here](https://www.kaggle.com/code/tarique7/lb-0-0240-h-m-ensemble-magic-multi-blend) as our final submission. And there was still one day left, so I decided to improve my ranking model based on my understanding, and see how it goes. \n\nAfter I added more features like the counts, popularity, and days, my ranking model improved. Then I added more recall methods from public notebooks (since I don't have enough time to build one myself) to generate more condidates, and train them on last 6 week instead of 1 week. The local cv improved a lot. Then I spent some time adding more simple count-based featuers. I can finish them in one day because I was actually in the right direction previously, and the rest of work was built on top of my previous version. \n\nSo I think I am just lucky to get the scores beyond my expectation.", "1784635": "edddddd \nAs a write up it is just chaotic but the crucial problem is that you didn't answer my questions.\n\nFirst - **which school** or institution is your \"class\" from?\n\nSecond - your **official teammates** should get a silver medal **for 1 respectively zero days** of activity on Kaggle. Is this your idea of fairness in Kaggle competitions?", "1784761": "many of them even dont have any submission" }, "source": "meta" }