s/he won a gold medal and ranked 19th placed with only 2 submissions aaand his/her last submission was 2 month ago.
\n\nMy next favorite: ziad, ranked 25th, with 5 submission and last submission 1 month ago.
\n\ncongratulations to both of you.
", "messages": [ { "id": "542470", "postDate": "06/04/2019 00:16:45", "content": "s/he won a gold medal and ranked 19th placed with only 2 submissions aaand his/her last submission was 2 month ago.
\n\nMy next favorite: ziad, ranked 25th, with 5 submission and last submission 1 month ago.
\n\ncongratulations to both of you.
", "rawMarkdown": "s/he won a gold medal and ranked 19th placed with only 2 submissions aaand his/her last submission was 2 month ago. \n\nMy next favorite: ziad, ranked 25th, with 5 submission and last submission 1 month ago. \n\ncongratulations to both of you.", "votes": null }, { "id": "542492", "postDate": "06/04/2019 00:26:53", "content": "maybe luck, maybe skills
", "rawMarkdown": "maybe luck, maybe skills", "votes": null }, { "id": "542507", "postDate": "06/04/2019 00:37:16", "content": "Most of us if only had 2-5 submissions probably would have been better off. A lot of effort for over-fitting the h... out of our models. But it was an interesting competition with most having medal level submissions that were not in their final selection.
", "rawMarkdown": "Most of us if only had 2-5 submissions probably would have been better off. A lot of effort for over-fitting the h... out of our models. But it was an interesting competition with most having medal level submissions that were not in their final selection.", "votes": null }, { "id": "542508", "postDate": "06/04/2019 00:38:11", "content": "Maybe both.
\n\nCongratulations to your team too. Youre among the few who survived the shake up. @amjad85 congratz to you too for tour first gold medal.
", "rawMarkdown": "Maybe both.\n\nCongratulations to your team too. Youre among the few who survived the shake up. @amjad85 congratz to you too for tour first gold medal.", "votes": null }, { "id": "542511", "postDate": "06/04/2019 00:40:35", "content": "It was a lesson that sticks though. No matter how much we read in books not to overfit, we need a lesson like this in the end.\nDafe badi ishalla ;)
", "rawMarkdown": "It was a lesson that sticks though. No matter how much we read in books not to overfit, we need a lesson like this in the end.\nDafe badi ishalla ;)", "votes": null }, { "id": "542513", "postDate": "06/04/2019 00:43:08", "content": "Ishalla ba ham ;)
", "rawMarkdown": "Ishalla ba ham ;)", "votes": null }, { "id": "542521", "postDate": "06/04/2019 00:49:43", "content": "Thank you. I wouldn't have survived without Giba
", "rawMarkdown": "Thank you. I wouldn't have survived without Giba", "votes": null }, { "id": "542524", "postDate": "06/04/2019 00:53:54", "content": "Congrats @amjad85 . Excellent job.
", "rawMarkdown": "Congrats @amjad85 . Excellent job.", "votes": null }, { "id": "542528", "postDate": "06/04/2019 00:55:54", "content": "frankly, thats another lesson i learned from this competition. I remember when you told him your model could benefit from his experience when you were at the top already. At that moment I thought maybe you can get a gold medal alone. But i was wrong and im glad that i was.
", "rawMarkdown": "frankly, thats another lesson i learned from this competition. I remember when you told him your model could benefit from his experience when you were at the top already. At that moment I thought maybe you can get a gold medal alone. But i was wrong and im glad that i was.", "votes": null }, { "id": "542697", "postDate": "06/04/2019 03:37:07", "content": "I think they should have a new rule, 50 submissions total each team
", "rawMarkdown": "I think they should have a new rule, 50 submissions total each team", "votes": null }, { "id": "542703", "postDate": "06/04/2019 03:44:55", "content": "Don't think it was overfitting. The test data set was different type. Any models (some of these were barely trained) got better CV and public LB scores than final top private LB scores.
", "rawMarkdown": "Don't think it was overfitting. The test data set was different type. Any models (some of these were barely trained) got better CV and public LB scores than final top private LB scores.", "votes": null }, { "id": "542709", "postDate": "06/04/2019 03:49:36", "content": "But ... Why?
", "rawMarkdown": "But ... Why?", "votes": null }, { "id": "542710", "postDate": "06/04/2019 03:51:30", "content": "@joejeo1 this competition had many problems but i learned a lot from it.
\n\nBtw, I won my very first competitions medal ^_^
", "rawMarkdown": "joejeo1 this competition had many problems but i learned a lot from it. \n\nBtw, I won my very first competitions medal ^_^", "votes": null }, { "id": "542733", "postDate": "06/04/2019 04:12:16", "content": "What's the point of submitting so many times with small numbers of public LB samples ?
", "rawMarkdown": "What's the point of submitting so many times with small numbers of public LB samples ?", "votes": null }, { "id": "542845", "postDate": "06/04/2019 06:27:43", "content": "Congrats
", "rawMarkdown": "Congrats", "votes": null }, { "id": "542936", "postDate": "06/04/2019 08:04:38", "content": "This is a stupid idea - most of us use Kaggle to learn - most of the learning is done AFTER the competition closes. One can go through all submissions and find out what worked and what didn't and then find out why. One can then use this knowledge in the future and not just for Kaggle.
\n\nThe only thing I would say is that I would love Kaggle to do a blind competition possibly on Kernels only - where we have zero knowledge of the test set if the data is tiny.
", "rawMarkdown": "This is a stupid idea - most of us use Kaggle to learn - most of the learning is done AFTER the competition closes. One can go through all submissions and find out what worked and what didn't and then find out why. One can then use this knowledge in the future and not just for Kaggle.\n\nThe only thing I would say is that I would love Kaggle to do a blind competition possibly on Kernels only - where we have zero knowledge of the test set if the data is tiny.", "votes": null }, { "id": "546782", "postDate": "06/06/2019 21:50:41", "content": "@mhviraf hopefully it is just the beginning, I see you are good at Instant Gratification comp, good luck there!
", "rawMarkdown": "mhviraf hopefully it is just the beginning, I see you are good at Instant Gratification comp, good luck there!", "votes": null }, { "id": "546788", "postDate": "06/06/2019 22:00:02", "content": "@zaharch Thanks. It was before someone ruins the competition. From now on, it's just luck there.
", "rawMarkdown": "zaharch Thanks. It was before someone ruins the competition. From now on, it's just luck there.", "votes": null } ], "comments": [ { "id": 542492, "author_name": "titericz", "author_url": "", "post_date": "06/04/2019 00:26:53", "content": "maybe luck, maybe skills
", "votes": null, "replies": [ { "id": 542508, "author_name": "mhviraf", "author_url": "", "post_date": "06/04/2019 00:38:11", "content": "Maybe both.
\n\nCongratulations to your team too. Youre among the few who survived the shake up. @amjad85 congratz to you too for tour first gold medal.
", "votes": null, "replies": [] }, { "id": 542521, "author_name": "amjad85", "author_url": "", "post_date": "06/04/2019 00:49:43", "content": "Thank you. I wouldn't have survived without Giba
", "votes": null, "replies": [] }, { "id": 542524, "author_name": "siavrez", "author_url": "", "post_date": "06/04/2019 00:53:54", "content": "Congrats @amjad85 . Excellent job.
", "votes": null, "replies": [] }, { "id": 542528, "author_name": "mhviraf", "author_url": "", "post_date": "06/04/2019 00:55:54", "content": "frankly, thats another lesson i learned from this competition. I remember when you told him your model could benefit from his experience when you were at the top already. At that moment I thought maybe you can get a gold medal alone. But i was wrong and im glad that i was.
", "votes": null, "replies": [] } ] }, { "id": 542507, "author_name": "siavrez", "author_url": "", "post_date": "06/04/2019 00:37:16", "content": "Most of us if only had 2-5 submissions probably would have been better off. A lot of effort for over-fitting the h... out of our models. But it was an interesting competition with most having medal level submissions that were not in their final selection.
", "votes": null, "replies": [ { "id": 542511, "author_name": "mhviraf", "author_url": "", "post_date": "06/04/2019 00:40:35", "content": "It was a lesson that sticks though. No matter how much we read in books not to overfit, we need a lesson like this in the end.\nDafe badi ishalla ;)
", "votes": null, "replies": [] }, { "id": 542513, "author_name": "siavrez", "author_url": "", "post_date": "06/04/2019 00:43:08", "content": "Ishalla ba ham ;)
", "votes": null, "replies": [] }, { "id": 542703, "author_name": "joejeo1", "author_url": "", "post_date": "06/04/2019 03:44:55", "content": "Don't think it was overfitting. The test data set was different type. Any models (some of these were barely trained) got better CV and public LB scores than final top private LB scores.
", "votes": null, "replies": [] }, { "id": 542710, "author_name": "mhviraf", "author_url": "", "post_date": "06/04/2019 03:51:30", "content": "@joejeo1 this competition had many problems but i learned a lot from it.
\n\nBtw, I won my very first competitions medal ^_^
", "votes": null, "replies": [] }, { "id": 542845, "author_name": "joejeo1", "author_url": "", "post_date": "06/04/2019 06:27:43", "content": "Congrats
", "votes": null, "replies": [] }, { "id": 546782, "author_name": "zaharch", "author_url": "", "post_date": "06/06/2019 21:50:41", "content": "@mhviraf hopefully it is just the beginning, I see you are good at Instant Gratification comp, good luck there!
", "votes": null, "replies": [] }, { "id": 546788, "author_name": "mhviraf", "author_url": "", "post_date": "06/06/2019 22:00:02", "content": "@zaharch Thanks. It was before someone ruins the competition. From now on, it's just luck there.
", "votes": null, "replies": [] } ] }, { "id": 542697, "author_name": "joejeo1", "author_url": "", "post_date": "06/04/2019 03:37:07", "content": "I think they should have a new rule, 50 submissions total each team
", "votes": null, "replies": [ { "id": 542709, "author_name": "mhviraf", "author_url": "", "post_date": "06/04/2019 03:49:36", "content": "But ... Why?
", "votes": null, "replies": [] }, { "id": 542733, "author_name": "joejeo1", "author_url": "", "post_date": "06/04/2019 04:12:16", "content": "What's the point of submitting so many times with small numbers of public LB samples ?
", "votes": null, "replies": [] }, { "id": 542936, "author_name": "scirpus", "author_url": "", "post_date": "06/04/2019 08:04:38", "content": "This is a stupid idea - most of us use Kaggle to learn - most of the learning is done AFTER the competition closes. One can go through all submissions and find out what worked and what didn't and then find out why. One can then use this knowledge in the future and not just for Kaggle.
\n\nThe only thing I would say is that I would love Kaggle to do a blind competition possibly on Kernels only - where we have zero knowledge of the test set if the data is tiny.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "542470": "s/he won a gold medal and ranked 19th placed with only 2 submissions aaand his/her last submission was 2 month ago. \n\nMy next favorite: ziad, ranked 25th, with 5 submission and last submission 1 month ago. \n\ncongratulations to both of you.", "542492": "maybe luck, maybe skills", "542507": "Most of us if only had 2-5 submissions probably would have been better off. A lot of effort for over-fitting the h... out of our models. But it was an interesting competition with most having medal level submissions that were not in their final selection.", "542508": "Maybe both.\n\nCongratulations to your team too. Youre among the few who survived the shake up. @amjad85 congratz to you too for tour first gold medal.", "542511": "It was a lesson that sticks though. No matter how much we read in books not to overfit, we need a lesson like this in the end.\nDafe badi ishalla ;)", "542513": "Ishalla ba ham ;)", "542521": "Thank you. I wouldn't have survived without Giba", "542524": "Congrats @amjad85 . Excellent job.", "542528": "frankly, thats another lesson i learned from this competition. I remember when you told him your model could benefit from his experience when you were at the top already. At that moment I thought maybe you can get a gold medal alone. But i was wrong and im glad that i was.", "542697": "I think they should have a new rule, 50 submissions total each team", "542703": "Don't think it was overfitting. The test data set was different type. Any models (some of these were barely trained) got better CV and public LB scores than final top private LB scores.", "542709": "But ... Why?", "542710": "joejeo1 this competition had many problems but i learned a lot from it. \n\nBtw, I won my very first competitions medal ^_^", "542733": "What's the point of submitting so many times with small numbers of public LB samples ?", "542845": "Congrats", "542936": "This is a stupid idea - most of us use Kaggle to learn - most of the learning is done AFTER the competition closes. One can go through all submissions and find out what worked and what didn't and then find out why. One can then use this knowledge in the future and not just for Kaggle.\n\nThe only thing I would say is that I would love Kaggle to do a blind competition possibly on Kernels only - where we have zero knowledge of the test set if the data is tiny.", "546782": "mhviraf hopefully it is just the beginning, I see you are good at Instant Gratification comp, good luck there!", "546788": "zaharch Thanks. It was before someone ruins the competition. From now on, it's just luck there." }, "source": "meta" }