Hi, Anyone using keras and having LB above 0.81 single model ?
", "messages": [ { "id": "612306", "postDate": "08/29/2019 17:59:17", "content": "Hi, Anyone using keras and having LB above 0.81 single model ?
", "rawMarkdown": "Hi, Anyone using keras and having LB above 0.81 single model ?", "votes": null }, { "id": "612871", "postDate": "08/30/2019 03:55:51", "content": "hi, i use tf.keras and my best single model (5 fold average) obtain 85.2 on LB
", "rawMarkdown": "hi, i use tf.keras and my best single model (5 fold average) obtain 85.2 on LB", "votes": null }, { "id": "612885", "postDate": "08/30/2019 04:07:39", "content": "wow
", "rawMarkdown": "wow", "votes": null }, { "id": "612890", "postDate": "08/30/2019 04:17:18", "content": "That's cool!
", "rawMarkdown": "That's cool!", "votes": null }, { "id": "612899", "postDate": "08/30/2019 04:36:01", "content": "To confirm that Keras is not inferior to Pytorch in this competition, my pure keras single model, having completely different training process than my great teammate @dathudeptrai \"Nguyen Quan Anh Minh\" also got .844 LB ...
\n\nI cannot give a training process details until the end of the competition, but if it manages to get a good Private score also [hopefully, this is true], I will write it up in details. (Already hinted a lot in my kernels & discussions though ;)
\n\nI answer this topic because in the previous competition, I usually asked this kind of question too :)
", "rawMarkdown": "To confirm that Keras is not inferior to Pytorch in this competition, my pure keras single model, having completely different training process than my great teammate @dathudeptrai \"Nguyen Quan Anh Minh\" also got .844 LB ... \n\nI cannot give a training process details until the end of the competition, but if it manages to get a good Private score also [hopefully, this is true], I will write it up in details. (Already hinted a lot in my kernels & discussions though ;)\n\nI answer this topic because in the previous competition, I usually asked this kind of question too :)", "votes": null }, { "id": "612906", "postDate": "08/30/2019 04:44:09", "content": "WOW! Lot to learn from you guys @ratthachat @dathudeptrai
", "rawMarkdown": "WOW! Lot to learn from you guys @ratthachat @dathudeptrai", "votes": null }, { "id": "612957", "postDate": "08/30/2019 05:43:11", "content": "wow +1
", "rawMarkdown": "wow +1", "votes": null }, { "id": "613006", "postDate": "08/30/2019 06:47:02", "content": "this also means you still did not ensemble :)
", "rawMarkdown": "this also means you still did not ensemble :)", "votes": null }, { "id": "613968", "postDate": "08/31/2019 01:26:42", "content": "hey, @garybios , I notice you improved a lot in recent days, can you share some tricks?
", "rawMarkdown": "hey, @garybios , I notice you improved a lot in recent days, can you share some tricks?", "votes": null }, { "id": "614093", "postDate": "08/31/2019 04:18:45", "content": "@garybios is already a very strong and highly experienced kaggler (much better than I am as I participated many competitions with him) … So, I am not surprised at all of his performance :)
", "rawMarkdown": "garybios is already a very strong and highly experienced kaggler (much better than I am as I participated many competitions with him) … So, I am not surprised at all of his performance :)", "votes": null }, { "id": "614113", "postDate": "08/31/2019 05:04:06", "content": "@hustkevin1037 @ratthachat I met some strange problems this competition, so the public lb may not be trusted.
", "rawMarkdown": "hustkevin1037 @ratthachat I met some strange problems this competition, so the public lb may not be trusted.", "votes": null }, { "id": "614251", "postDate": "08/31/2019 08:57:14", "content": "@garybios what problems? Am i wrong if I say that you mainly improved with ensemble? Also did you find useful any of the tricks others shared here in discussion?
", "rawMarkdown": "garybios what problems? Am i wrong if I say that you mainly improved with ensemble? Also did you find useful any of the tricks others shared here in discussion?", "votes": null }, { "id": "614394", "postDate": "08/31/2019 12:37:02", "content": "@garybios what's your highest performance of single model?
", "rawMarkdown": "garybios what's your highest performance of single model?", "votes": null }, { "id": "615405", "postDate": "09/01/2019 22:05:06", "content": "Is there any substantial difference between keras and tf.keras?
", "rawMarkdown": "Is there any substantial difference between keras and tf.keras?", "votes": null } ], "comments": [ { "id": 612871, "author_name": "dathudeptrai", "author_url": "", "post_date": "08/30/2019 03:55:51", "content": "hi, i use tf.keras and my best single model (5 fold average) obtain 85.2 on LB
", "votes": null, "replies": [ { "id": 612885, "author_name": "buaazijian", "author_url": "", "post_date": "08/30/2019 04:07:39", "content": "wow
", "votes": null, "replies": [] }, { "id": 612890, "author_name": "seefun", "author_url": "", "post_date": "08/30/2019 04:17:18", "content": "That's cool!
", "votes": null, "replies": [] }, { "id": 612906, "author_name": "virajbagal", "author_url": "", "post_date": "08/30/2019 04:44:09", "content": "WOW! Lot to learn from you guys @ratthachat @dathudeptrai
", "votes": null, "replies": [] }, { "id": 612957, "author_name": "garybios", "author_url": "", "post_date": "08/30/2019 05:43:11", "content": "wow +1
", "votes": null, "replies": [] }, { "id": 613006, "author_name": "valanm", "author_url": "", "post_date": "08/30/2019 06:47:02", "content": "this also means you still did not ensemble :)
", "votes": null, "replies": [] }, { "id": 613968, "author_name": "hustkevin1037", "author_url": "", "post_date": "08/31/2019 01:26:42", "content": "hey, @garybios , I notice you improved a lot in recent days, can you share some tricks?
", "votes": null, "replies": [] }, { "id": 614093, "author_name": "ratthachat", "author_url": "", "post_date": "08/31/2019 04:18:45", "content": "@garybios is already a very strong and highly experienced kaggler (much better than I am as I participated many competitions with him) … So, I am not surprised at all of his performance :)
", "votes": null, "replies": [] }, { "id": 614113, "author_name": "garybios", "author_url": "", "post_date": "08/31/2019 05:04:06", "content": "@hustkevin1037 @ratthachat I met some strange problems this competition, so the public lb may not be trusted.
", "votes": null, "replies": [] }, { "id": 614251, "author_name": "valanm", "author_url": "", "post_date": "08/31/2019 08:57:14", "content": "@garybios what problems? Am i wrong if I say that you mainly improved with ensemble? Also did you find useful any of the tricks others shared here in discussion?
", "votes": null, "replies": [] }, { "id": 614394, "author_name": "hustkevin1037", "author_url": "", "post_date": "08/31/2019 12:37:02", "content": "@garybios what's your highest performance of single model?
", "votes": null, "replies": [] }, { "id": 615405, "author_name": "masterm4", "author_url": "", "post_date": "09/01/2019 22:05:06", "content": "Is there any substantial difference between keras and tf.keras?
", "votes": null, "replies": [] } ] }, { "id": 612899, "author_name": "ratthachat", "author_url": "", "post_date": "08/30/2019 04:36:01", "content": "To confirm that Keras is not inferior to Pytorch in this competition, my pure keras single model, having completely different training process than my great teammate @dathudeptrai \"Nguyen Quan Anh Minh\" also got .844 LB ...
\n\nI cannot give a training process details until the end of the competition, but if it manages to get a good Private score also [hopefully, this is true], I will write it up in details. (Already hinted a lot in my kernels & discussions though ;)
\n\nI answer this topic because in the previous competition, I usually asked this kind of question too :)
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "612306": "Hi, Anyone using keras and having LB above 0.81 single model ?", "612871": "hi, i use tf.keras and my best single model (5 fold average) obtain 85.2 on LB", "612885": "wow", "612890": "That's cool!", "612899": "To confirm that Keras is not inferior to Pytorch in this competition, my pure keras single model, having completely different training process than my great teammate @dathudeptrai \"Nguyen Quan Anh Minh\" also got .844 LB ... \n\nI cannot give a training process details until the end of the competition, but if it manages to get a good Private score also [hopefully, this is true], I will write it up in details. (Already hinted a lot in my kernels & discussions though ;)\n\nI answer this topic because in the previous competition, I usually asked this kind of question too :)", "612906": "WOW! Lot to learn from you guys @ratthachat @dathudeptrai", "612957": "wow +1", "613006": "this also means you still did not ensemble :)", "613968": "hey, @garybios , I notice you improved a lot in recent days, can you share some tricks?", "614093": "garybios is already a very strong and highly experienced kaggler (much better than I am as I participated many competitions with him) … So, I am not surprised at all of his performance :)", "614113": "hustkevin1037 @ratthachat I met some strange problems this competition, so the public lb may not be trusted.", "614251": "garybios what problems? Am i wrong if I say that you mainly improved with ensemble? Also did you find useful any of the tricks others shared here in discussion?", "614394": "garybios what's your highest performance of single model?", "615405": "Is there any substantial difference between keras and tf.keras?" }, "source": "meta" }