Hi everyone. Thanks to all of you for showing interest in the competition. As mentioned in the description, your algorithms can help create tools that will ultimately assist in creating new medicines, and I hope that this challenge will be fun for all of us and that you end up creating a wide range of interesting solutions.
\n\nOnly a day in and people have already been very helpful and active on the discussion board and I hope that this continues. Will Gerard and I will try to answer your questions in an ongoing fashion, but we want to avoid shoehorning you too much towards specific approaches to this challenge. However, if something is unclear, worded weirdly or possibly incorrect, we will try to react promptly.
\n\nFeel free to post any questions, comments, suggestions or feedback in this thread!
", "messages": [ { "id": "539856", "postDate": "05/30/2019 16:04:34", "content": "Hi everyone. Thanks to all of you for showing interest in the competition. As mentioned in the description, your algorithms can help create tools that will ultimately assist in creating new medicines, and I hope that this challenge will be fun for all of us and that you end up creating a wide range of interesting solutions.
\n\nOnly a day in and people have already been very helpful and active on the discussion board and I hope that this continues. Will Gerard and I will try to answer your questions in an ongoing fashion, but we want to avoid shoehorning you too much towards specific approaches to this challenge. However, if something is unclear, worded weirdly or possibly incorrect, we will try to react promptly.
\n\nFeel free to post any questions, comments, suggestions or feedback in this thread!
", "rawMarkdown": "Hi everyone. Thanks to all of you for showing interest in the competition. As mentioned in the description, your algorithms can help create tools that will ultimately assist in creating new medicines, and I hope that this challenge will be fun for all of us and that you end up creating a wide range of interesting solutions. \n\nOnly a day in and people have already been very helpful and active on the discussion board and I hope that this continues. [Will Gerard](https://www.kaggle.com/willgerrard) and I will try to answer your questions in an ongoing fashion, but we want to avoid shoehorning you too much towards specific approaches to this challenge. However, if something is unclear, worded weirdly or possibly incorrect, we will try to react promptly.\n\nFeel free to post any questions, comments, suggestions or feedback in this thread!", "votes": null }, { "id": "540489", "postDate": "05/31/2019 15:15:50", "content": "Hi,
\n\nI have read that we are not allowed to use the QM datasets as a source of external information.\nIs the information in the QM datasets only contained in these datasets? What is to keep someone from extracting this information from a secondary source? I assume that will be checked? But what if they extract that information unintentionally? Thanks.
", "rawMarkdown": "Hi,\n\nI have read that we are not allowed to use the QM datasets as a source of external information.\nIs the information in the QM datasets only contained in these datasets? What is to keep someone from extracting this information from a secondary source? I assume that will be checked? But what if they extract that information unintentionally? Thanks.", "votes": null }, { "id": "540503", "postDate": "05/31/2019 15:32:33", "content": "The information in the QM7 and QM9 data sets is as far as I'm aware an unique source for this data. To be eligible to win a prize, the code that generated the predictions will be checked to make sure that no prohibited external data has been used.\nAll external data has to be announced in the appropriate thread, so we will have a chance to respond if this breaks the rules or not. In general, we should be able to train the model on a completely different set of molecules without having to run quantum mechanics or similarly expensive computations, and the external data can't prohibit academic or industrial use of the methodology.
", "rawMarkdown": "The information in the QM7 and QM9 data sets is as far as I'm aware an unique source for this data. To be eligible to win a prize, the code that generated the predictions will be checked to make sure that no prohibited external data has been used.\nAll external data has to be announced in the appropriate thread, so we will have a chance to respond if this breaks the rules or not. In general, we should be able to train the model on a completely different set of molecules without having to run quantum mechanics or similarly expensive computations, and the external data can't prohibit academic or industrial use of the methodology.", "votes": null }, { "id": "546053", "postDate": "06/06/2019 07:55:53", "content": "I consider participating in the competition but I have the same question as above. If it is possible to obtain the test data target values then between different approaches and random seeds one can select the best scoring one, and it is not detectable by looking at the model code as you mentioned. How this will be prevented?
\n\nYou mentioned \"In general, we should be able to train the model on a completely different set of molecules\" but it is not part of the planned evaluation process right now, is it?
\n\nThank you!
", "rawMarkdown": "I consider participating in the competition but I have the same question as above. If it is possible to obtain the test data target values then between different approaches and random seeds one can select the best scoring one, and it is not detectable by looking at the model code as you mentioned. How this will be prevented?\n\nYou mentioned \"In general, we should be able to train the model on a completely different set of molecules\" but it is not part of the planned evaluation process right now, is it?\n\nThank you!", "votes": null }, { "id": "550218", "postDate": "06/11/2019 12:16:32", "content": "Thanks for hosting as well.
", "rawMarkdown": "Thanks for hosting as well.", "votes": null }, { "id": "568091", "postDate": "07/04/2019 11:41:00", "content": "Hello Lars,\nThank you for hosting this competition and participating in the discussion forum.\nI have recently developed an algorithm to expedite the formation of the Density Matrix, which can be further used to calculate various properties and speed up other Quantum Chemical calculations.\nAppreciate if you could please share your contact details for further discussion.\nBest Regards.
", "rawMarkdown": "Hello Lars,\nThank you for hosting this competition and participating in the discussion forum.\nI have recently developed an algorithm to expedite the formation of the Density Matrix, which can be further used to calculate various properties and speed up other Quantum Chemical calculations.\nAppreciate if you could please share your contact details for further discussion.\nBest Regards.", "votes": null }, { "id": "569083", "postDate": "07/06/2019 02:57:20", "content": "Hi Lars,\nThe training data has the corresponding additional data files, but they are not available for the test data. Is there anyway to generate these files for the test data?
", "rawMarkdown": "Hi Lars,\nThe training data has the corresponding additional data files, but they are not available for the test data. Is there anyway to generate these files for the test data?", "votes": null }, { "id": "576760", "postDate": "07/16/2019 01:21:27", "content": "Hi,
\n\ncould you clarify in which coordinate frames the dipole moments and magnetic shielding tensors are specified?
\n\nAs observed by @shahrdar in this thread: dipole moments, it appears that the dipole moments are not defined using the same reference frame as the atom positions.
", "rawMarkdown": "Hi,\n\ncould you clarify in which coordinate frames the dipole moments and magnetic shielding tensors are specified?\n\nAs observed by @shahrdar in this thread: [dipole moments](https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836), it appears that the dipole moments are not defined using the same reference frame as the atom positions.", "votes": null } ], "comments": [ { "id": 540489, "author_name": "tblack", "author_url": "", "post_date": "05/31/2019 15:15:50", "content": "Hi,
\n\nI have read that we are not allowed to use the QM datasets as a source of external information.\nIs the information in the QM datasets only contained in these datasets? What is to keep someone from extracting this information from a secondary source? I assume that will be checked? But what if they extract that information unintentionally? Thanks.
", "votes": null, "replies": [ { "id": 540503, "author_name": "larsbratholm", "author_url": "", "post_date": "05/31/2019 15:32:33", "content": "The information in the QM7 and QM9 data sets is as far as I'm aware an unique source for this data. To be eligible to win a prize, the code that generated the predictions will be checked to make sure that no prohibited external data has been used.\nAll external data has to be announced in the appropriate thread, so we will have a chance to respond if this breaks the rules or not. In general, we should be able to train the model on a completely different set of molecules without having to run quantum mechanics or similarly expensive computations, and the external data can't prohibit academic or industrial use of the methodology.
", "votes": null, "replies": [] }, { "id": 546053, "author_name": "zaharch", "author_url": "", "post_date": "06/06/2019 07:55:53", "content": "I consider participating in the competition but I have the same question as above. If it is possible to obtain the test data target values then between different approaches and random seeds one can select the best scoring one, and it is not detectable by looking at the model code as you mentioned. How this will be prevented?
\n\nYou mentioned \"In general, we should be able to train the model on a completely different set of molecules\" but it is not part of the planned evaluation process right now, is it?
\n\nThank you!
", "votes": null, "replies": [] } ] }, { "id": 550218, "author_name": "stephenmugisha", "author_url": "", "post_date": "06/11/2019 12:16:32", "content": "Thanks for hosting as well.
", "votes": null, "replies": [] }, { "id": 568091, "author_name": "qubitninja", "author_url": "", "post_date": "07/04/2019 11:41:00", "content": "Hello Lars,\nThank you for hosting this competition and participating in the discussion forum.\nI have recently developed an algorithm to expedite the formation of the Density Matrix, which can be further used to calculate various properties and speed up other Quantum Chemical calculations.\nAppreciate if you could please share your contact details for further discussion.\nBest Regards.
", "votes": null, "replies": [] }, { "id": 569083, "author_name": "manasghosal", "author_url": "", "post_date": "07/06/2019 02:57:20", "content": "Hi Lars,\nThe training data has the corresponding additional data files, but they are not available for the test data. Is there anyway to generate these files for the test data?
", "votes": null, "replies": [] }, { "id": 576760, "author_name": "cedben", "author_url": "", "post_date": "07/16/2019 01:21:27", "content": "Hi,
\n\ncould you clarify in which coordinate frames the dipole moments and magnetic shielding tensors are specified?
\n\nAs observed by @shahrdar in this thread: dipole moments, it appears that the dipole moments are not defined using the same reference frame as the atom positions.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "539856": "Hi everyone. Thanks to all of you for showing interest in the competition. As mentioned in the description, your algorithms can help create tools that will ultimately assist in creating new medicines, and I hope that this challenge will be fun for all of us and that you end up creating a wide range of interesting solutions. \n\nOnly a day in and people have already been very helpful and active on the discussion board and I hope that this continues. [Will Gerard](https://www.kaggle.com/willgerrard) and I will try to answer your questions in an ongoing fashion, but we want to avoid shoehorning you too much towards specific approaches to this challenge. However, if something is unclear, worded weirdly or possibly incorrect, we will try to react promptly.\n\nFeel free to post any questions, comments, suggestions or feedback in this thread!", "540489": "Hi,\n\nI have read that we are not allowed to use the QM datasets as a source of external information.\nIs the information in the QM datasets only contained in these datasets? What is to keep someone from extracting this information from a secondary source? I assume that will be checked? But what if they extract that information unintentionally? Thanks.", "540503": "The information in the QM7 and QM9 data sets is as far as I'm aware an unique source for this data. To be eligible to win a prize, the code that generated the predictions will be checked to make sure that no prohibited external data has been used.\nAll external data has to be announced in the appropriate thread, so we will have a chance to respond if this breaks the rules or not. In general, we should be able to train the model on a completely different set of molecules without having to run quantum mechanics or similarly expensive computations, and the external data can't prohibit academic or industrial use of the methodology.", "546053": "I consider participating in the competition but I have the same question as above. If it is possible to obtain the test data target values then between different approaches and random seeds one can select the best scoring one, and it is not detectable by looking at the model code as you mentioned. How this will be prevented?\n\nYou mentioned \"In general, we should be able to train the model on a completely different set of molecules\" but it is not part of the planned evaluation process right now, is it?\n\nThank you!", "550218": "Thanks for hosting as well.", "568091": "Hello Lars,\nThank you for hosting this competition and participating in the discussion forum.\nI have recently developed an algorithm to expedite the formation of the Density Matrix, which can be further used to calculate various properties and speed up other Quantum Chemical calculations.\nAppreciate if you could please share your contact details for further discussion.\nBest Regards.", "569083": "Hi Lars,\nThe training data has the corresponding additional data files, but they are not available for the test data. Is there anyway to generate these files for the test data?", "576760": "Hi,\n\ncould you clarify in which coordinate frames the dipole moments and magnetic shielding tensors are specified?\n\nAs observed by @shahrdar in this thread: [dipole moments](https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836), it appears that the dipole moments are not defined using the same reference frame as the atom positions." }, "source": "meta" }