First, congratulations to all the participants, the MoleculeLearning's (such a great pun :p) teammates Harshit Mehta, Mukharbek Organokov, Felipe Mello, and Stanislav Blinov, and the top participants.
\n\nYou all did great! And a special thank you to the organizers for the well crafted dataset and competition.
\n\nI am still waiting for the 3 top solutions to learn new things (if they would share their solutions of course). :p
\n\nAnyway, that was an amazing competition and probably the best I have ever entered in Kaggle (I haven't done many so I surely have missed other great competitions).
\n\nWhat made it exceptional? As with exceptional things, a lot of parts where great and these added up to make an exceptional experience. Here are some:
\n\nNow, some bits that I've learned during this competition:
\n\nA funny anecdote before concluding: during the last night, one of our team mates managed to increase our CV score while I was sleeping (had a rough week :p). Prior to sleeping, I selected two submissions (to avoid bad surprises if there was any shakeup). Well, we learned that since I am the \"team lead\", I was the only one allowed to select the submissions. The good thing is that this didn't have a huge impact and we only lost 4 places. Catastrophe avoided. :D
\n\nAgain, congratulations!
", "messages": [ { "id": "611386", "postDate": "08/29/2019 08:30:45", "content": "First, congratulations to all the participants, the MoleculeLearning's (such a great pun :p) teammates Harshit Mehta, Mukharbek Organokov, Felipe Mello, and Stanislav Blinov, and the top participants.
\n\nYou all did great! And a special thank you to the organizers for the well crafted dataset and competition.
\n\nI am still waiting for the 3 top solutions to learn new things (if they would share their solutions of course). :p
\n\nAnyway, that was an amazing competition and probably the best I have ever entered in Kaggle (I haven't done many so I surely have missed other great competitions).
\n\nWhat made it exceptional? As with exceptional things, a lot of parts where great and these added up to make an exceptional experience. Here are some:
\n\nNow, some bits that I've learned during this competition:
\n\nA funny anecdote before concluding: during the last night, one of our team mates managed to increase our CV score while I was sleeping (had a rough week :p). Prior to sleeping, I selected two submissions (to avoid bad surprises if there was any shakeup). Well, we learned that since I am the \"team lead\", I was the only one allowed to select the submissions. The good thing is that this didn't have a huge impact and we only lost 4 places. Catastrophe avoided. :D
\n\nAgain, congratulations!
", "rawMarkdown": "First, **congratulations** to all the participants, the MoleculeLearning's (such a great pun :p) teammates [Harshit Mehta](https://www.kaggle.com/harshit92), [Mukharbek Organokov](https://www.kaggle.com/muhakabartay), [Felipe Mello](https://www.kaggle.com/felipemello), and [Stanislav Blinov](https://www.kaggle.com/stanislavblinov), and the top participants. \n\nYou all did great! And a special thank you to the organizers for the well crafted dataset and competition.\n\nI am still waiting for the 3 top solutions to learn new things (if they would share their solutions of course). :p\n\nAnyway, that was an amazing competition and probably the best I have ever entered in Kaggle (I haven't done many so I surely have missed other great competitions). \n\nWhat made it exceptional? As with exceptional things, a lot of parts where great and these added up to make an exceptional experience. Here are some:\n\n1. **LB/CV/Private high agreement**: this doesn't happen in all competitions and it often ruins the fun. Well done to the organizers for such a balanced dataset.\n2. **Teaming**: this was my first time teaming up and boy, I was missing a lot of fun. Waiting to do it again soon!\n3. **Great discussions and kernels**: I won't cite any since there are many but the overall discussions and shared kernels were very informative and helpful. Thanks to all!\n\n\nNow, some bits that I've learned during this competition: \n\n1. **Reading** the forums and running the kernels is very useful!\n2. **GNN** models seem to be working very well: never tried any before and have learned a lot about them. I am excited to try these on upcoming challenges.\n3. A lot of **molecular chemistry concepts**: I have never been good in chemistry but managed to learn a lot of new things (maybe because molecular chemistry is closer to physics than traditional chemistry?).\n4. Tried various **schedulers** and **optimizers**: [ReduceLROnPlateau](https://pytorch.org/docs/stable/_modules/torch/optim/lr_scheduler.html) and [RAdam](https://github.com/LiyuanLucasLiu/RAdam) to name a few.\n5. Tried **deep learning frameworks**: pytorch (which I have used to some extent before) and the really new one to me was [chainer](https://chainer.org). \n\nA funny anecdote before concluding: during the last night, one of our team mates managed to increase our CV score while I was sleeping (had a rough week :p). Prior to sleeping, I selected two submissions (to avoid bad surprises if there was any shakeup). Well, we learned that since I am the \"team lead\", I was the only one allowed to select the submissions. The good thing is that this didn't have a huge impact and we only lost 4 places. Catastrophe avoided. :D\n\nAgain, congratulations!", "votes": null }, { "id": "611550", "postDate": "08/29/2019 10:14:12", "content": "Last but not the least,\nImportance of observing the data, specially this discussion (thanks for initiating it):\nhttps://www.kaggle.com/c/champs-scalar-coupling/discussion/101644
\n\nCongratulation!
", "rawMarkdown": "Last but not the least,\nImportance of observing the data, specially this discussion (thanks for initiating it):\nhttps://www.kaggle.com/c/champs-scalar-coupling/discussion/101644\n\nCongratulation!", "votes": null }, { "id": "611565", "postDate": "08/29/2019 10:21:08", "content": "One of my teammates ( @pcjimmmy ) ask me to make some preliminary final selections before driving home from work, in case I got in an accident. :-) I think that's a good policy!
", "rawMarkdown": "One of my teammates ( @pcjimmmy ) ask me to make some preliminary final selections before driving home from work, in case I got in an accident. :-) I think that's a good policy!", "votes": null } ], "comments": [ { "id": 611550, "author_name": "krishnakantsingh14", "author_url": "", "post_date": "08/29/2019 10:14:12", "content": "Last but not the least,\nImportance of observing the data, specially this discussion (thanks for initiating it):\nhttps://www.kaggle.com/c/champs-scalar-coupling/discussion/101644
\n\nCongratulation!
", "votes": null, "replies": [] }, { "id": 611565, "author_name": "abendorf", "author_url": "", "post_date": "08/29/2019 10:21:08", "content": "One of my teammates ( @pcjimmmy ) ask me to make some preliminary final selections before driving home from work, in case I got in an accident. :-) I think that's a good policy!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "611386": "First, **congratulations** to all the participants, the MoleculeLearning's (such a great pun :p) teammates [Harshit Mehta](https://www.kaggle.com/harshit92), [Mukharbek Organokov](https://www.kaggle.com/muhakabartay), [Felipe Mello](https://www.kaggle.com/felipemello), and [Stanislav Blinov](https://www.kaggle.com/stanislavblinov), and the top participants. \n\nYou all did great! And a special thank you to the organizers for the well crafted dataset and competition.\n\nI am still waiting for the 3 top solutions to learn new things (if they would share their solutions of course). :p\n\nAnyway, that was an amazing competition and probably the best I have ever entered in Kaggle (I haven't done many so I surely have missed other great competitions). \n\nWhat made it exceptional? As with exceptional things, a lot of parts where great and these added up to make an exceptional experience. Here are some:\n\n1. **LB/CV/Private high agreement**: this doesn't happen in all competitions and it often ruins the fun. Well done to the organizers for such a balanced dataset.\n2. **Teaming**: this was my first time teaming up and boy, I was missing a lot of fun. Waiting to do it again soon!\n3. **Great discussions and kernels**: I won't cite any since there are many but the overall discussions and shared kernels were very informative and helpful. Thanks to all!\n\n\nNow, some bits that I've learned during this competition: \n\n1. **Reading** the forums and running the kernels is very useful!\n2. **GNN** models seem to be working very well: never tried any before and have learned a lot about them. I am excited to try these on upcoming challenges.\n3. A lot of **molecular chemistry concepts**: I have never been good in chemistry but managed to learn a lot of new things (maybe because molecular chemistry is closer to physics than traditional chemistry?).\n4. Tried various **schedulers** and **optimizers**: [ReduceLROnPlateau](https://pytorch.org/docs/stable/_modules/torch/optim/lr_scheduler.html) and [RAdam](https://github.com/LiyuanLucasLiu/RAdam) to name a few.\n5. Tried **deep learning frameworks**: pytorch (which I have used to some extent before) and the really new one to me was [chainer](https://chainer.org). \n\nA funny anecdote before concluding: during the last night, one of our team mates managed to increase our CV score while I was sleeping (had a rough week :p). Prior to sleeping, I selected two submissions (to avoid bad surprises if there was any shakeup). Well, we learned that since I am the \"team lead\", I was the only one allowed to select the submissions. The good thing is that this didn't have a huge impact and we only lost 4 places. Catastrophe avoided. :D\n\nAgain, congratulations!", "611550": "Last but not the least,\nImportance of observing the data, specially this discussion (thanks for initiating it):\nhttps://www.kaggle.com/c/champs-scalar-coupling/discussion/101644\n\nCongratulation!", "611565": "One of my teammates ( @pcjimmmy ) ask me to make some preliminary final selections before driving home from work, in case I got in an accident. :-) I think that's a good policy!" }, "source": "meta" }