Apologies to those I injured! :/
", "messages": [ { "id": "1335263", "postDate": "06/04/2021 05:44:50", "content": "Apologies to those I injured! :/
", "rawMarkdown": "Apologies to those I injured! :/", "votes": null }, { "id": "1335272", "postDate": "06/04/2021 06:00:09", "content": "If someone dig into the discussion, they can find there are some hints about that. So, just go ahead and challenge to another competition.
", "rawMarkdown": "If someone dig into the discussion, they can find there are some hints about that. So, just go ahead and challenge to another competition.", "votes": null }, { "id": "1335503", "postDate": "06/04/2021 09:04:32", "content": "I think because it was shared late, people didn't like it. If it was a weak earlier, nobody would have a problem with that. It looked like also that you only shared it because you didn't reach a gold place :)
", "rawMarkdown": "I think because it was shared late, people didn't like it. If it was a weak earlier, nobody would have a problem with that. It looked like also that you only shared it because you didn't reach a gold place :)", "votes": null }, { "id": "1335560", "postDate": "06/04/2021 09:28:10", "content": "I wrote the discussion itself almost immediately after I found out this trick. I was hyped and I thought that I could get the gold.
\nI thought that the discussion was clear enough, but then it turned out it wasn't.
It's true that when I wrote the specifics, I was almost certain that I'll have no gold. I'm ashamed, but this had a little to do with the fact that I shared the specifics, I think.
\nBut the real thing that made me write it was that I wasn't so sure that it works for everybody, and before it turns out that for someone it doesn't, they waste time on figuring out what I said, too..
\nSo I wrote what I did and how it affected my CV score so that everyone can judge for themselves if this is really the route they want to take in their last days or not, so that I don't waste anybody's time with information that I'm not 100% sure will help..
I'm not sure if I would've shared it if I was in the gold zone. I guess I would've in that case, too, but not sure.
\nBut in the future I won't share anything so influential in the last couple of days, that one thing is for sure.
By the way, you did pretty well with a single 3090 :) I bet most of top teams have a lot of computing resources.
", "rawMarkdown": "By the way, you did pretty well with a single 3090 :) I bet most of top teams have a lot of computing resources.", "votes": null }, { "id": "1335579", "postDate": "06/04/2021 09:39:43", "content": "Thank you!
\nI used my 1080Ti, too. It consumed 2/3 the power of 3090 but only gave 1/4 of compute. Terrible efficiency, but had to use everything I got :D
@tugstugi I knew something has changed!
\nYou're a GM now, Congratulations!! :))
We really appreciate and admire your sharing in this competition, especially the normalization of Inchi. Although It didn't improve our models much (the prediction target of most of our models was SMILES), It shocked us. I and the other two members of my team all have chemical backgrounds. We are masters or doctoral students majoring in drug design. But previously, we only knew that one molecular structure could correspond to multiple smiles and always thought that the Inchi corresponding to a molecular structure was unique.
", "rawMarkdown": "We really appreciate and admire your sharing in this competition, especially the normalization of Inchi. Although It didn't improve our models much (the prediction target of most of our models was SMILES), It shocked us. I and the other two members of my team all have chemical backgrounds. We are masters or doctoral students majoring in drug design. But previously, we only knew that one molecular structure could correspond to multiple smiles and always thought that the Inchi corresponding to a molecular structure was unique.", "votes": null }, { "id": "1335725", "postDate": "06/04/2021 11:55:37", "content": "There is only one heroism in the world：to see the world as it is and to love it. nofreewill is the person like this. The basic philosophy of the development of artificial intelligence is sharing and communication. I am very happy to meet such a hero who has implemented philosophy.
", "rawMarkdown": "There is only one heroism in the world：to see the world as it is and to love it. nofreewill is the person like this. The basic philosophy of the development of artificial intelligence is sharing and communication. I am very happy to meet such a hero who has implemented philosophy.", "votes": null }, { "id": "1335760", "postDate": "06/04/2021 12:21:35", "content": "@zhongfeisheng6666 @jiachengxiong thank you very much for your (exaggeratedly:D) kind words!
\nAlso, your score is astonishing! If I took in my validation the lowest LD prediction from all of my models, that gave me 0.49.. and you made 0.51 at test time.. Hats off!
\nCongratulations on winning!
@nofreewill thanks for being so persistent in pushing the importance of normalisation, without that I wouldn't have found @wuliaokaola's implementation of your ideas. Even though I ended up outside the medals, that last 0.11 meant that I ended up satisfied with the best score I could manage.
", "rawMarkdown": "nofreewill thanks for being so persistent in pushing the importance of normalisation, without that I wouldn't have found @wuliaokaola's implementation of your ideas. Even though I ended up outside the medals, that last 0.11 meant that I ended up satisfied with the best score I could manage.", "votes": null }, { "id": "1336460", "postDate": "06/05/2021 00:10:16", "content": "Perhaps you can sell your 1080Ti now. Here the current resale price of a used 1080Ti is 80% of the price I paid for a new one four years ago!
", "rawMarkdown": "Perhaps you can sell your 1080Ti now. Here the current resale price of a used 1080Ti is 80% of the price I paid for a new one *four years ago*!", "votes": null }, { "id": "1336463", "postDate": "06/05/2021 00:13:04", "content": "Validation of InChI had been discussed several times during the months of this competition. The only difference was 1) @nofreewill quantified the improvement he saw, and 2) as he has previously shared useful ideas people took his suggestions seriously.
", "rawMarkdown": "Validation of InChI had been discussed several times during the months of this competition. The only difference was 1) @nofreewill quantified the improvement he saw, and 2) as he has previously shared useful ideas people took his suggestions seriously.", "votes": null }, { "id": "1336467", "postDate": "06/05/2021 00:23:37", "content": "I am not sure what you referred to, normalize or validation filter?. Though I didn't use normalize_inchi() to improve my LB due to other reason, I learned it. Thanks for sharing.
", "rawMarkdown": "I am not sure what you referred to, normalize or validation filter?. Though I didn't use normalize_inchi() to improve my LB due to other reason, I learned it. Thanks for sharing.", "votes": null }, { "id": "1336812", "postDate": "06/05/2021 08:19:27", "content": "No worries and congratz for the good result!
\nI believe you'll get yourself a solo gold later or sooner ;)
You weren't even that far from a medal. :)
\nThank you for the kind words!
Thank you! Kaggle is amazing, so I'm sure I'll fight for it in future competitions :))
", "rawMarkdown": "Thank you! Kaggle is amazing, so I'm sure I'll fight for it in future competitions :))", "votes": null }, { "id": "1336827", "postDate": "06/05/2021 08:29:47", "content": "No need to apologize. There is no free will, you had to do it.
", "rawMarkdown": "No need to apologize. There is no free will, you had to do it.", "votes": null }, { "id": "1336831", "postDate": "06/05/2021 08:31:51", "content": "I think I'll keep the 1080Ti, but for Kaggle I'll use it only for quick ideas. Also, it's nice to have it for other things to do while the 3090 is sweating in the other machine. 1),2); good points.
", "rawMarkdown": "I think I'll keep the 1080Ti, but for Kaggle I'll use it only for quick ideas. Also, it's nice to have it for other things to do while the 3090 is sweating in the other machine. 1),2); good points.", "votes": null }, { "id": "1336837", "postDate": "06/05/2021 08:33:22", "content": "Had to, the same way I had to apologize… :D
", "rawMarkdown": "Had to, the same way I had to apologize... :D", "votes": null }, { "id": "1337170", "postDate": "06/05/2021 13:13:08", "content": "To be honest, if I had your level of insight about this competition, I would have rent gpu/tpu severs to help improve LB.
", "rawMarkdown": "To be honest, if I had your level of insight about this competition, I would have rent gpu/tpu severs to help improve LB.", "votes": null }, { "id": "1337196", "postDate": "06/05/2021 13:22:57", "content": "I was thinking about that at the start. I took a look on prices, and found them too high after how much money I already spent. And as I really thought that I could get a gold with only what I have, I decided that I won't spend more money.
\nBut after it became more and more unlikely that I can get the gold, the idea of renting just did not cross my mind… what if it did, probably I could have done better. who knows :D
", "rawMarkdown": "I was thinking about that at the start. I took a look on prices, and found them too high after how much money I already spent. And as I really thought that I could get a gold with only what I have, I decided that I won't spend more money.\n\nBut after it became more and more unlikely that I can get the gold, the idea of renting just did not cross my mind... what if it did, probably I could have done better. who knows :D", "votes": null }, { "id": "1337278", "postDate": "06/05/2021 14:23:50", "content": "No worries, I think everyone learned lesson =)
\nBy the way thank you for sharing the normalization script. It seems like a lot of people used it including our team =)
Thanks! and Welcome! :))
", "rawMarkdown": "Thanks! and Welcome! :))", "votes": null }, { "id": "1338721", "postDate": "06/06/2021 16:53:27", "content": "It would be really nice if you could share your solution. I am really interested in seeing the 1st place solution :)
", "rawMarkdown": "It would be really nice if you could share your solution. I am really interested in seeing the 1st place solution :)", "votes": null } ], "comments": [ { "id": 1335272, "author_name": "shinewine", "author_url": "", "post_date": "06/04/2021 06:00:09", "content": "If someone dig into the discussion, they can find there are some hints about that. So, just go ahead and challenge to another competition.
", "votes": null, "replies": [] }, { "id": 1335503, "author_name": "tugstugi", "author_url": "", "post_date": "06/04/2021 09:04:32", "content": "I think because it was shared late, people didn't like it. If it was a weak earlier, nobody would have a problem with that. It looked like also that you only shared it because you didn't reach a gold place :)
", "votes": null, "replies": [ { "id": 1335560, "author_name": "nofreewill", "author_url": "", "post_date": "06/04/2021 09:28:10", "content": "I wrote the discussion itself almost immediately after I found out this trick. I was hyped and I thought that I could get the gold.
\nI thought that the discussion was clear enough, but then it turned out it wasn't.
It's true that when I wrote the specifics, I was almost certain that I'll have no gold. I'm ashamed, but this had a little to do with the fact that I shared the specifics, I think.
\nBut the real thing that made me write it was that I wasn't so sure that it works for everybody, and before it turns out that for someone it doesn't, they waste time on figuring out what I said, too..
\nSo I wrote what I did and how it affected my CV score so that everyone can judge for themselves if this is really the route they want to take in their last days or not, so that I don't waste anybody's time with information that I'm not 100% sure will help..
I'm not sure if I would've shared it if I was in the gold zone. I guess I would've in that case, too, but not sure.
\nBut in the future I won't share anything so influential in the last couple of days, that one thing is for sure.
By the way, you did pretty well with a single 3090 :) I bet most of top teams have a lot of computing resources.
", "votes": null, "replies": [] }, { "id": 1335579, "author_name": "nofreewill", "author_url": "", "post_date": "06/04/2021 09:39:43", "content": "Thank you!
\nI used my 1080Ti, too. It consumed 2/3 the power of 3090 but only gave 1/4 of compute. Terrible efficiency, but had to use everything I got :D
@tugstugi I knew something has changed!
\nYou're a GM now, Congratulations!! :))
Perhaps you can sell your 1080Ti now. Here the current resale price of a used 1080Ti is 80% of the price I paid for a new one four years ago!
", "votes": null, "replies": [] }, { "id": 1336463, "author_name": "talktocharles", "author_url": "", "post_date": "06/05/2021 00:13:04", "content": "Validation of InChI had been discussed several times during the months of this competition. The only difference was 1) @nofreewill quantified the improvement he saw, and 2) as he has previously shared useful ideas people took his suggestions seriously.
", "votes": null, "replies": [] }, { "id": 1336831, "author_name": "nofreewill", "author_url": "", "post_date": "06/05/2021 08:31:51", "content": "I think I'll keep the 1080Ti, but for Kaggle I'll use it only for quick ideas. Also, it's nice to have it for other things to do while the 3090 is sweating in the other machine. 1),2); good points.
", "votes": null, "replies": [] }, { "id": 1337170, "author_name": "dragonzhang", "author_url": "", "post_date": "06/05/2021 13:13:08", "content": "To be honest, if I had your level of insight about this competition, I would have rent gpu/tpu severs to help improve LB.
", "votes": null, "replies": [] }, { "id": 1337196, "author_name": "nofreewill", "author_url": "", "post_date": "06/05/2021 13:22:57", "content": "I was thinking about that at the start. I took a look on prices, and found them too high after how much money I already spent. And as I really thought that I could get a gold with only what I have, I decided that I won't spend more money.
\nBut after it became more and more unlikely that I can get the gold, the idea of renting just did not cross my mind… what if it did, probably I could have done better. who knows :D
", "votes": null, "replies": [] } ] }, { "id": 1335719, "author_name": "jiachengxiong", "author_url": "", "post_date": "06/04/2021 11:53:09", "content": "We really appreciate and admire your sharing in this competition, especially the normalization of Inchi. Although It didn't improve our models much (the prediction target of most of our models was SMILES), It shocked us. I and the other two members of my team all have chemical backgrounds. We are masters or doctoral students majoring in drug design. But previously, we only knew that one molecular structure could correspond to multiple smiles and always thought that the Inchi corresponding to a molecular structure was unique.
", "votes": null, "replies": [] }, { "id": 1335725, "author_name": "zhongfeisheng6666", "author_url": "", "post_date": "06/04/2021 11:55:37", "content": "There is only one heroism in the world：to see the world as it is and to love it. nofreewill is the person like this. The basic philosophy of the development of artificial intelligence is sharing and communication. I am very happy to meet such a hero who has implemented philosophy.
", "votes": null, "replies": [ { "id": 1338721, "author_name": "eakdag", "author_url": "", "post_date": "06/06/2021 16:53:27", "content": "It would be really nice if you could share your solution. I am really interested in seeing the 1st place solution :)
", "votes": null, "replies": [] } ] }, { "id": 1335760, "author_name": "nofreewill", "author_url": "", "post_date": "06/04/2021 12:21:35", "content": "@zhongfeisheng6666 @jiachengxiong thank you very much for your (exaggeratedly:D) kind words!
\nAlso, your score is astonishing! If I took in my validation the lowest LD prediction from all of my models, that gave me 0.49.. and you made 0.51 at test time.. Hats off!
\nCongratulations on winning!
@nofreewill thanks for being so persistent in pushing the importance of normalisation, without that I wouldn't have found @wuliaokaola's implementation of your ideas. Even though I ended up outside the medals, that last 0.11 meant that I ended up satisfied with the best score I could manage.
", "votes": null, "replies": [ { "id": 1336820, "author_name": "nofreewill", "author_url": "", "post_date": "06/05/2021 08:23:39", "content": "You weren't even that far from a medal. :)
\nThank you for the kind words!
I am not sure what you referred to, normalize or validation filter?. Though I didn't use normalize_inchi() to improve my LB due to other reason, I learned it. Thanks for sharing.
", "votes": null, "replies": [] }, { "id": 1336812, "author_name": "haqishen", "author_url": "", "post_date": "06/05/2021 08:19:27", "content": "No worries and congratz for the good result!
\nI believe you'll get yourself a solo gold later or sooner ;)
Thank you! Kaggle is amazing, so I'm sure I'll fight for it in future competitions :))
", "votes": null, "replies": [] } ] }, { "id": 1336827, "author_name": "aerdem4", "author_url": "", "post_date": "06/05/2021 08:29:47", "content": "No need to apologize. There is no free will, you had to do it.
", "votes": null, "replies": [ { "id": 1336837, "author_name": "nofreewill", "author_url": "", "post_date": "06/05/2021 08:33:22", "content": "Had to, the same way I had to apologize… :D
", "votes": null, "replies": [] } ] }, { "id": 1337278, "author_name": "drhabib", "author_url": "", "post_date": "06/05/2021 14:23:50", "content": "No worries, I think everyone learned lesson =)
\nBy the way thank you for sharing the normalization script. It seems like a lot of people used it including our team =)
Thanks! and Welcome! :))
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1335263": "Apologies to those I injured! :/", "1335272": "If someone dig into the discussion, they can find there are some hints about that. So, just go ahead and challenge to another competition.", "1335503": "I think because it was shared late, people didn't like it. If it was a weak earlier, nobody would have a problem with that. It looked like also that you only shared it because you didn't reach a gold place :)", "1335560": "I wrote the discussion itself almost immediately after I found out this trick. I was hyped and I thought that I could get the gold.\nI thought that the discussion was clear enough, but then it turned out it wasn't.\n\nIt's true that when I wrote the specifics, **I was almost certain that I'll have no gold**. I'm ashamed, but this had a little to do with the fact that I shared the specifics, I think.\n\nBut the real thing that made me write it was that I wasn't so sure that it works for everybody, and before it turns out that for someone it doesn't, they waste time on figuring out what I said, too..\nSo I wrote what I did and how it affected my CV score so that everyone can judge for themselves if this is really the route they want to take in their last days or not, so that I don't waste anybody's time with information that I'm not 100% sure will help..\n\nI'm not sure if I would've shared it if I was in the gold zone. I guess I would've in that case, too, but not sure.\nBut in the future I won't share anything so influential in the last couple of days, that one thing is for sure.", "1335572": "By the way, you did pretty well with a single 3090 :) I bet most of top teams have a lot of computing resources.", "1335579": "Thank you!\nI used my 1080Ti, too. It consumed 2/3 the power of 3090 but only gave 1/4 of compute. Terrible efficiency, but had to use everything I got :D", "1335640": "tugstugi I knew something has changed!\nYou're a GM now, Congratulations!! :))", "1335719": "We really appreciate and admire your sharing in this competition, especially the normalization of Inchi. Although It didn't improve our models much (the prediction target of most of our models was SMILES), It shocked us. I and the other two members of my team all have chemical backgrounds. We are masters or doctoral students majoring in drug design. But previously, we only knew that one molecular structure could correspond to multiple smiles and always thought that the Inchi corresponding to a molecular structure was unique.", "1335725": "There is only one heroism in the world：to see the world as it is and to love it. nofreewill is the person like this. The basic philosophy of the development of artificial intelligence is sharing and communication. I am very happy to meet such a hero who has implemented philosophy.", "1335760": "zhongfeisheng6666 @jiachengxiong thank you very much for your (exaggeratedly:D) kind words!\nAlso, your score is astonishing! If I took in my validation the lowest LD prediction from all of my models, that gave me 0.49.. and you made 0.51 at test time.. Hats off!\nCongratulations on winning!", "1336412": "nofreewill thanks for being so persistent in pushing the importance of normalisation, without that I wouldn't have found @wuliaokaola's implementation of your ideas. Even though I ended up outside the medals, that last 0.11 meant that I ended up satisfied with the best score I could manage.", "1336460": "Perhaps you can sell your 1080Ti now. Here the current resale price of a used 1080Ti is 80% of the price I paid for a new one *four years ago*!", "1336463": "Validation of InChI had been discussed several times during the months of this competition. The only difference was 1) @nofreewill quantified the improvement he saw, and 2) as he has previously shared useful ideas people took his suggestions seriously.", "1336467": "I am not sure what you referred to, normalize or validation filter?. Though I didn't use normalize_inchi() to improve my LB due to other reason, I learned it. Thanks for sharing.", "1336812": "No worries and congratz for the good result!\nI believe you'll get yourself a solo gold later or sooner ;)", "1336820": "You weren't even that far from a medal. :)\nThank you for the kind words!", "1336823": "Thank you! Kaggle is amazing, so I'm sure I'll fight for it in future competitions :))", "1336827": "No need to apologize. There is no free will, you had to do it.", "1336831": "I think I'll keep the 1080Ti, but for Kaggle I'll use it only for quick ideas. Also, it's nice to have it for other things to do while the 3090 is sweating in the other machine. 1),2); good points.", "1336837": "Had to, the same way I had to apologize... :D", "1337170": "To be honest, if I had your level of insight about this competition, I would have rent gpu/tpu severs to help improve LB.", "1337196": "I was thinking about that at the start. I took a look on prices, and found them too high after how much money I already spent. And as I really thought that I could get a gold with only what I have, I decided that I won't spend more money.\n\nBut after it became more and more unlikely that I can get the gold, the idea of renting just did not cross my mind... what if it did, probably I could have done better. who knows :D", "1337278": "No worries, I think everyone learned lesson =)\nBy the way thank you for sharing the normalization script. It seems like a lot of people used it including our team =)", "1338038": "Thanks! and Welcome! :))", "1338721": "It would be really nice if you could share your solution. I am really interested in seeing the 1st place solution :)" }, "source": "meta" }