In this competition, you will be using molecular structure information to predict the magnetic interaction between atoms in molecules. The key to doing this well is engineering features from the base structural information.
\n\nTo get you started, you can take a look at the Atomic Distance Benchmark, which uses a single engineered feature.
\n\nPlease read the competition pages carefully; this competition has a significant amount of useful background information on the Overview, Evaluation, Data, etc., pages.
\n\nHave fun and good luck!
", "messages": [ { "id": 539279, "postDate": "2019-05-29T20:13:51.477Z", "content": "In this competition, you will be using molecular structure information to predict the magnetic interaction between atoms in molecules. The key to doing this well is engineering features from the base structural information.
\n\nTo get you started, you can take a look at the Atomic Distance Benchmark, which uses a single engineered feature.
\n\nPlease read the competition pages carefully; this competition has a significant amount of useful background information on the Overview, Evaluation, Data, etc., pages.
\n\nHave fun and good luck!
", "rawMarkdown": "In this competition, you will be using molecular structure information to predict the magnetic interaction between atoms in molecules. The key to doing this well is engineering features from the base structural information.\n\nTo get you started, you can take a look at the [Atomic Distance Benchmark](https://www.kaggle.com/inversion/atomic-distance-benchmark/), which uses a single engineered feature.\n\nPlease read the competition pages carefully; this competition has a significant amount of useful background information on the Overview, Evaluation, Data, etc., pages.\n\nHave fun and good luck!", "votes": 41 }, { "id": 539475, "postDate": "2019-05-30T05:08:03.300Z", "content": "After reading overview page , I thought \"Scalar coupling constant\" is a continuous variable but in evaluation page : \nFor each id in the test set, you must predict a probability for the scalar_coupling_constant variable.
Probability ?
", "rawMarkdown": "After reading overview page , I thought \"Scalar coupling constant\" is a continuous variable but in evaluation page : \n``` For each id in the test set, you must predict a probability for the scalar_coupling_constant variable. ```\n\nProbability ? ", "votes": 9, "replies": [ { "id": 539751, "postDate": "2019-05-30T13:30:23.387Z", "content": "Good catch. The scalar coupling constants are continuous variables and any notion of probability didn't make sense. I've updated the page.
", "rawMarkdown": "Good catch. The scalar coupling constants are continuous variables and any notion of probability didn't make sense. I've updated the page.", "votes": 8 } ] }, { "id": 539300, "postDate": "2019-05-29T21:31:39.440Z", "content": "Hello,\nsince all four components of scalar coupling values are given I assume they are results of QC calculations, correct?\nIf so, were they calculated in vacuo or rather there was solvent presence modelled aroud the molecules?\nAlso, if they were calculated are they result of ab initio methods (HF, MP2, CC, ..) or rather DFT?
", "rawMarkdown": "Hello,\nsince all four components of scalar coupling values are given I assume they are results of QC calculations, correct?\nIf so, were they calculated in vacuo or rather there was solvent presence modelled aroud the molecules?\nAlso, if they were calculated are they result of ab initio methods (HF, MP2, CC, ..) or rather DFT?", "votes": 8 }, { "id": 1014634, "postDate": "2020-09-17T16:04:55.993Z", "content": "@inversion
\nIt appears that the Late Submission functionality is broken. It doesn't agree with the number of rows in the sample_submission file nor with the test.csv file.
Evaluation Exception: Submission must have 2505542 rows
", "rawMarkdown": "@inversion \nIt appears that the Late Submission functionality is broken. It doesn't agree with the number of rows in the sample_submission file nor with the test.csv file. \n\n**Evaluation Exception: Submission must have 2505542 rows**", "votes": 1, "replies": [ { "id": 1014771, "postDate": "2020-09-17T17:24:18.903Z", "content": "Thanks for flagging. I'll look into the cause.
", "rawMarkdown": "Thanks for flagging. I'll look into the cause.", "votes": 2 } ] }, { "id": 539505, "postDate": "2019-05-30T06:10:29.600Z", "content": "Thank you for pointing out the crux of the problem.
", "rawMarkdown": "Thank you for pointing out the crux of the problem.", "votes": 3 }, { "id": 540902, "postDate": "2019-06-01T10:22:02.783Z", "content": "Dear Sir,\n From Structure==> additional info==> scalar constant, do you mean to say we can use the strutural info of the molecules to create a model for some additional info which may be related to the scalar constant and in turn use those additional info to find the scalar constant? Am I right?
", "rawMarkdown": "Dear Sir,\n From Structure==> additional info==> scalar constant, do you mean to say we can use the strutural info of the molecules to create a model for some additional info which may be related to the scalar constant and in turn use those additional info to find the scalar constant? Am I right?", "votes": 1, "replies": [ { "id": 541916, "postDate": "2019-06-03T07:40:58.113Z", "content": "That's how I understood it, too.
", "rawMarkdown": "That's how I understood it, too.", "votes": 1 }, { "id": 543232, "postDate": "2019-06-04T11:57:32.100Z", "content": "cool.
", "rawMarkdown": "cool." } ] }, { "id": 540885, "postDate": "2019-06-01T09:49:45.960Z", "content": "Dear Mr. Lars,\n I am new to kaggle competition. I need some clarification regarding the expectation from your side. I read the project overview and it seems clear but I am little lost with possible output of the project. I understand from the overview that you are expecting to calculate the scalar constant. Is it that you are expecting to find the scalar cobstant as a function of addition data sets and molecular structure or only in terms of molecular structure. In the test file you have mentioned the atom id and the type of molecule so the out put has to be like you input the type of molecule and the atom id and you are expecting the scalar constant as an output? Thank you.
", "rawMarkdown": "Dear Mr. Lars,\n I am new to kaggle competition. I need some clarification regarding the expectation from your side. I read the project overview and it seems clear but I am little lost with possible output of the project. I understand from the overview that you are expecting to calculate the scalar constant. Is it that you are expecting to find the scalar cobstant as a function of addition data sets and molecular structure or only in terms of molecular structure. In the test file you have mentioned the atom id and the type of molecule so the out put has to be like you input the type of molecule and the atom id and you are expecting the scalar constant as an output? Thank you.", "votes": 1, "replies": [ { "id": 541101, "postDate": "2019-06-01T18:11:56.377Z", "content": "An example of the submission format is available in the sample_submission.csv file. You can either just use the molecule coordinates, atom types and scalar coupling type to make your predictions, or you can make additional use of tools that can provide some extra information for an arbitrary molecule (e.g. information about chemical bonds with openbabel). You might also want to create your own tools/models from the supplementary data available for the training set to improve your predictions.
How is atom_index index assigned? Can you please comment on its role from the \"physics\" perspective?
It's just an index. Any permutation of the atom order would still give the same physical observables.
", "rawMarkdown": "It's just an index. Any permutation of the atom order would still give the same physical observables." } ] }, { "id": 539363, "postDate": "2019-05-30T01:28:48.980Z", "content": "What is the point in providing additional information for training molecules only?
", "rawMarkdown": "What is the point in providing additional information for training molecules only?", "votes": 2, "replies": [ { "id": 539889, "postDate": "2019-05-30T16:56:55.853Z", "content": "The point of the competition is to calculate the target from the structural information. But if, say, you find that some of that additional information is connected to the target, it might be useful to create a model pipeline like this:
\n\nstructure => (additional info) => target
", "rawMarkdown": "The point of the competition is to calculate the target from the structural information. But if, say, you find that some of that additional information is connected to the target, it might be useful to create a model pipeline like this:\n\nstructure => (additional info) => target", "votes": 2 }, { "id": 541520, "postDate": "2019-06-02T15:49:28.300Z", "content": "This question might be a bit naive so ignore my ignorance.
\n\nHow can we predict the target if a model trained on additional information along with training data set is tested on a test data set without same set of additional information?
", "rawMarkdown": "This question might be a bit naive so ignore my ignorance. \n\nHow can we predict the target if a model trained on additional information along with training data set is tested on a test data set without same set of additional information?", "votes": 3 }, { "id": 541915, "postDate": "2019-06-03T07:40:28.730Z", "content": "@vikasmalhotra08 What he is suggesting is that you might want to build two models in a pipeline. The first model would predict the additional features from the original data as an intermediate step. Afterwards, you would use both the structural data and your predicted features in your final model. So basically you would create the information that is missing yourself.
", "rawMarkdown": "@vikasmalhotra08 What he is suggesting is that you *might* want to build two models in a pipeline. The first model would predict the additional features from the original data as an intermediate step. Afterwards, you would use both the structural data and your predicted features in your final model. So basically you would create the information that is missing yourself.", "votes": 7 }, { "id": 541926, "postDate": "2019-06-03T07:52:45.773Z", "content": "I've got the same question too.
\n\nThe only information we can get from test data is molecular name, atom 1, atom 2 and type.\nNow that train set and test set share the same atoms and type, we may generate some extra useful information from atom 1/atom 2/type at test time, which we extract from the train set and additional files provided. May be atom_x + atom_y + type_k have a probility of p to have e potential energy, something like that.
\n\nBut i'm really not in the least sure if I comprehend inversion's idea the right way = =!
", "rawMarkdown": "I've got the same question too.\n\nThe only information we can get from test data is molecular name, atom 1, atom 2 and type.\nNow that train set and test set share the same atoms and type, we may generate some extra useful information from atom 1/atom 2/type at test time, which we extract from the train set and additional files provided. May be atom_x + atom_y + type_k have a probility of p to have e potential energy, something like that.\n\nBut i'm really not in the least sure if I comprehend inversion's idea the right way = =!" }, { "id": 542087, "postDate": "2019-06-03T12:52:19.670Z", "content": "@lesamu Ah, gotcha, something like transfer learning approach, it didn't hit me before. Thanks for pointing out potential meaning of the statement.
", "rawMarkdown": "@lesamu Ah, gotcha, something like transfer learning approach, it didn't hit me before. Thanks for pointing out potential meaning of the statement. " }, { "id": 542295, "postDate": "2019-06-03T17:01:13.227Z", "content": "Thanks a lot, I've been looking for this hint, wasn't sure I understood the point of this additional data before.
", "rawMarkdown": "Thanks a lot, I've been looking for this hint, wasn't sure I understood the point of this additional data before. ", "votes": 1 }, { "id": 569326, "postDate": "2019-07-06T13:06:59.360Z", "content": "Very nice hint, thanks
", "rawMarkdown": "Very nice hint, thanks" } ] }, { "id": 541751, "postDate": "2019-06-03T01:36:29.407Z", "content": "Are the coupling constants in the dataset empirical coupling constants measured by NMR, or are they the results of some theoretical calculation with multiple (zero Kelvin, no solvent, DFT, etc. etc.) assumptions?
", "rawMarkdown": "Are the coupling constants in the dataset empirical coupling constants measured by NMR, or are they the results of some theoretical calculation with multiple (zero Kelvin, no solvent, DFT, etc. etc.) assumptions?" }, { "id": 648655, "postDate": "2019-10-14T13:04:15.940Z", "content": "I have little expertise in this domain, but I wonder what's the point of predicting scalar couplings given the 3-D structure of a molecule. As I know we can use NMR to gain insight into a molecule’s structure with the help of scalar couplings. But if we don't know the 3-D structure, how can we predict the scalar couplings. On the contrary, if we have known the 3D structure, there is no need to do this prediction. So I don't know what's the meaning of this competition task.
", "rawMarkdown": "I have little expertise in this domain, but I wonder what's the point of predicting scalar couplings given the 3-D structure of a molecule. As I know we can use NMR to gain insight into a molecule’s structure with the help of scalar couplings. But if we don't know the 3-D structure, how can we predict the scalar couplings. On the contrary, if we have known the 3D structure, there is no need to do this prediction. So I don't know what's the meaning of this competition task." }, { "id": 561985, "postDate": "2019-06-26T22:57:36.063Z", "content": "I tried to run each cell of this kernel, but the cell related to \"folds\" is taking too much time(more than half an hour). Now I just commit the kernel but the processing is still going on. Could you please tell me how to figure out this problem.
", "rawMarkdown": "I tried to run each cell of this kernel, but the cell related to \"folds\" is taking too much time(more than half an hour). Now I just commit the kernel but the processing is still going on. Could you please tell me how to figure out this problem." }, { "id": 539443, "postDate": "2019-05-30T03:45:38.307Z", "content": "the level of theory should be provided for the potential energy calculations.
", "rawMarkdown": "the level of theory should be provided for the potential energy calculations." }, { "id": 560850, "postDate": "2019-06-25T21:30:06.443Z", "rawMarkdown": "", "isDeleted": true }, { "id": 554094, "postDate": "2019-06-17T00:47:37.937Z", "rawMarkdown": "", "isDeleted": true }, { "id": 560330, "postDate": "2019-06-25T09:58:32.367Z", "content": "Thanks!
", "rawMarkdown": "Thanks!" }, { "id": 554217, "postDate": "2019-06-17T06:41:31.153Z", "content": "Thanks !
", "rawMarkdown": "Thanks !" }, { "id": 553522, "postDate": "2019-06-15T21:25:10.283Z", "content": "Many thanks!
", "rawMarkdown": "Many thanks!" }, { "id": 547627, "postDate": "2019-06-08T01:27:50.370Z", "content": "Thanks.
", "rawMarkdown": "Thanks." } ], "comments": [ { "id": 539475, "author_name": "redshellspy", "author_url": "", "post_date": "2019-05-30T05:08:03.300000", "content": "After reading overview page , I thought \"Scalar coupling constant\" is a continuous variable but in evaluation page : \nFor each id in the test set, you must predict a probability for the scalar_coupling_constant variable.
Probability ?
", "votes": 9, "replies": [ { "id": 539751, "author_name": "Lars Bratholm", "author_url": "", "post_date": "2019-05-30T13:30:23.387000", "content": "Good catch. The scalar coupling constants are continuous variables and any notion of probability didn't make sense. I've updated the page.
", "votes": 8, "replies": [] } ] }, { "id": 539300, "author_name": "Display Name", "author_url": "", "post_date": "2019-05-29T21:31:39.440000", "content": "Hello,\nsince all four components of scalar coupling values are given I assume they are results of QC calculations, correct?\nIf so, were they calculated in vacuo or rather there was solvent presence modelled aroud the molecules?\nAlso, if they were calculated are they result of ab initio methods (HF, MP2, CC, ..) or rather DFT?
", "votes": 8, "replies": [] }, { "id": 1014634, "author_name": "David Patton", "author_url": "", "post_date": "2020-09-17T16:04:55.993000", "content": "@inversion
\nIt appears that the Late Submission functionality is broken. It doesn't agree with the number of rows in the sample_submission file nor with the test.csv file.
Evaluation Exception: Submission must have 2505542 rows
", "votes": 1, "replies": [ { "id": 1014771, "author_name": "inversion", "author_url": "", "post_date": "2020-09-17T17:24:18.903000", "content": "Thanks for flagging. I'll look into the cause.
", "votes": 2, "replies": [] } ] }, { "id": 539505, "author_name": "Eric R", "author_url": "", "post_date": "2019-05-30T06:10:29.600000", "content": "Thank you for pointing out the crux of the problem.
", "votes": 3, "replies": [] }, { "id": 540902, "author_name": "Ganesh Selvaraj", "author_url": "", "post_date": "2019-06-01T10:22:02.783000", "content": "Dear Sir,\n From Structure==> additional info==> scalar constant, do you mean to say we can use the strutural info of the molecules to create a model for some additional info which may be related to the scalar constant and in turn use those additional info to find the scalar constant? Am I right?
", "votes": 1, "replies": [ { "id": 541916, "author_name": "Samuel", "author_url": "", "post_date": "2019-06-03T07:40:58.113000", "content": "That's how I understood it, too.
", "votes": 1, "replies": [] }, { "id": 543232, "author_name": "Ganesh Selvaraj", "author_url": "", "post_date": "2019-06-04T11:57:32.100000", "content": "cool.
", "votes": 0, "replies": [] } ] }, { "id": 540885, "author_name": "Ganesh Selvaraj", "author_url": "", "post_date": "2019-06-01T09:49:45.960000", "content": "Dear Mr. Lars,\n I am new to kaggle competition. I need some clarification regarding the expectation from your side. I read the project overview and it seems clear but I am little lost with possible output of the project. I understand from the overview that you are expecting to calculate the scalar constant. Is it that you are expecting to find the scalar cobstant as a function of addition data sets and molecular structure or only in terms of molecular structure. In the test file you have mentioned the atom id and the type of molecule so the out put has to be like you input the type of molecule and the atom id and you are expecting the scalar constant as an output? Thank you.
", "votes": 1, "replies": [ { "id": 541101, "author_name": "Lars Bratholm", "author_url": "", "post_date": "2019-06-01T18:11:56.377000", "content": "An example of the submission format is available in the sample_submission.csv file. You can either just use the molecule coordinates, atom types and scalar coupling type to make your predictions, or you can make additional use of tools that can provide some extra information for an arbitrary molecule (e.g. information about chemical bonds with openbabel). You might also want to create your own tools/models from the supplementary data available for the training set to improve your predictions.
How is atom_index index assigned? Can you please comment on its role from the \"physics\" perspective?
It's just an index. Any permutation of the atom order would still give the same physical observables.
", "votes": 0, "replies": [] } ] }, { "id": 539363, "author_name": "Binil George", "author_url": "", "post_date": "2019-05-30T01:28:48.980000", "content": "What is the point in providing additional information for training molecules only?
", "votes": 2, "replies": [ { "id": 539889, "author_name": "inversion", "author_url": "", "post_date": "2019-05-30T16:56:55.853000", "content": "The point of the competition is to calculate the target from the structural information. But if, say, you find that some of that additional information is connected to the target, it might be useful to create a model pipeline like this:
\n\nstructure => (additional info) => target
", "votes": 2, "replies": [] }, { "id": 541520, "author_name": "Nebula", "author_url": "", "post_date": "2019-06-02T15:49:28.300000", "content": "This question might be a bit naive so ignore my ignorance.
\n\nHow can we predict the target if a model trained on additional information along with training data set is tested on a test data set without same set of additional information?
", "votes": 3, "replies": [] }, { "id": 541915, "author_name": "Samuel", "author_url": "", "post_date": "2019-06-03T07:40:28.730000", "content": "@vikasmalhotra08 What he is suggesting is that you might want to build two models in a pipeline. The first model would predict the additional features from the original data as an intermediate step. Afterwards, you would use both the structural data and your predicted features in your final model. So basically you would create the information that is missing yourself.
", "votes": 7, "replies": [] }, { "id": 541926, "author_name": "littlebeandog", "author_url": "", "post_date": "2019-06-03T07:52:45.773000", "content": "I've got the same question too.
\n\nThe only information we can get from test data is molecular name, atom 1, atom 2 and type.\nNow that train set and test set share the same atoms and type, we may generate some extra useful information from atom 1/atom 2/type at test time, which we extract from the train set and additional files provided. May be atom_x + atom_y + type_k have a probility of p to have e potential energy, something like that.
\n\nBut i'm really not in the least sure if I comprehend inversion's idea the right way = =!
", "votes": 0, "replies": [] }, { "id": 542087, "author_name": "Nebula", "author_url": "", "post_date": "2019-06-03T12:52:19.670000", "content": "@lesamu Ah, gotcha, something like transfer learning approach, it didn't hit me before. Thanks for pointing out potential meaning of the statement.
", "votes": 0, "replies": [] }, { "id": 542295, "author_name": "Tim Mironov", "author_url": "", "post_date": "2019-06-03T17:01:13.227000", "content": "Thanks a lot, I've been looking for this hint, wasn't sure I understood the point of this additional data before.
", "votes": 1, "replies": [] }, { "id": 569326, "author_name": "Joshua Olulana", "author_url": "", "post_date": "2019-07-06T13:06:59.360000", "content": "Very nice hint, thanks
", "votes": 0, "replies": [] } ] }, { "id": 541751, "author_name": "double-d", "author_url": "", "post_date": "2019-06-03T01:36:29.407000", "content": "Are the coupling constants in the dataset empirical coupling constants measured by NMR, or are they the results of some theoretical calculation with multiple (zero Kelvin, no solvent, DFT, etc. etc.) assumptions?
", "votes": 0, "replies": [] }, { "id": 648655, "author_name": "Caiqing Jian", "author_url": "", "post_date": "2019-10-14T13:04:15.940000", "content": "I have little expertise in this domain, but I wonder what's the point of predicting scalar couplings given the 3-D structure of a molecule. As I know we can use NMR to gain insight into a molecule’s structure with the help of scalar couplings. But if we don't know the 3-D structure, how can we predict the scalar couplings. On the contrary, if we have known the 3D structure, there is no need to do this prediction. So I don't know what's the meaning of this competition task.
", "votes": 0, "replies": [] }, { "id": 561985, "author_name": "Muhammad", "author_url": "", "post_date": "2019-06-26T22:57:36.063000", "content": "I tried to run each cell of this kernel, but the cell related to \"folds\" is taking too much time(more than half an hour). Now I just commit the kernel but the processing is still going on. Could you please tell me how to figure out this problem.
", "votes": 0, "replies": [] }, { "id": 539443, "author_name": "Erik Carl Vik", "author_url": "", "post_date": "2019-05-30T03:45:38.307000", "content": "the level of theory should be provided for the potential energy calculations.
", "votes": 0, "replies": [] }, { "id": 560850, "author_name": "", "author_url": "", "post_date": "2019-06-25T21:30:06.443000", "content": "", "votes": 0, "replies": [] }, { "id": 554094, "author_name": "", "author_url": "", "post_date": "2019-06-17T00:47:37.937000", "content": "", "votes": 0, "replies": [] }, { "id": 560330, "author_name": "Antonio", "author_url": "", "post_date": "2019-06-25T09:58:32.367000", "content": "Thanks!
", "votes": 0, "replies": [] }, { "id": 554217, "author_name": "Bhavik bansal", "author_url": "", "post_date": "2019-06-17T06:41:31.153000", "content": "Thanks !
", "votes": 0, "replies": [] }, { "id": 553522, "author_name": "Rishiraj Raghu", "author_url": "", "post_date": "2019-06-15T21:25:10.283000", "content": "Many thanks!
", "votes": 0, "replies": [] }, { "id": 547627, "author_name": "miyaco2kmz", "author_url": "", "post_date": "2019-06-08T01:27:50.370000", "content": "Thanks.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "539279": "In this competition, you will be using molecular structure information to predict the magnetic interaction between atoms in molecules. The key to doing this well is engineering features from the base structural information.\n\nTo get you started, you can take a look at the [Atomic Distance Benchmark](https://www.kaggle.com/inversion/atomic-distance-benchmark/), which uses a single engineered feature.\n\nPlease read the competition pages carefully; this competition has a significant amount of useful background information on the Overview, Evaluation, Data, etc., pages.\n\nHave fun and good luck!", "539475": "After reading overview page , I thought \"Scalar coupling constant\" is a continuous variable but in evaluation page : \n``` For each id in the test set, you must predict a probability for the scalar_coupling_constant variable. ```\n\nProbability ? ", "539300": "Hello,\nsince all four components of scalar coupling values are given I assume they are results of QC calculations, correct?\nIf so, were they calculated in vacuo or rather there was solvent presence modelled aroud the molecules?\nAlso, if they were calculated are they result of ab initio methods (HF, MP2, CC, ..) or rather DFT?", "1014634": "@inversion \nIt appears that the Late Submission functionality is broken. It doesn't agree with the number of rows in the sample_submission file nor with the test.csv file. \n\n**Evaluation Exception: Submission must have 2505542 rows**", "539505": "Thank you for pointing out the crux of the problem.", "540902": "Dear Sir,\n From Structure==> additional info==> scalar constant, do you mean to say we can use the strutural info of the molecules to create a model for some additional info which may be related to the scalar constant and in turn use those additional info to find the scalar constant? Am I right?", "540885": "Dear Mr. Lars,\n I am new to kaggle competition. I need some clarification regarding the expectation from your side. I read the project overview and it seems clear but I am little lost with possible output of the project. I understand from the overview that you are expecting to calculate the scalar constant. Is it that you are expecting to find the scalar cobstant as a function of addition data sets and molecular structure or only in terms of molecular structure. In the test file you have mentioned the atom id and the type of molecule so the out put has to be like you input the type of molecule and the atom id and you are expecting the scalar constant as an output? Thank you.", "552115": "How is `atom_index` index assigned? Can you please comment on its role from the \"physics\" perspective? ", "539363": "What is the point in providing additional information for training molecules only?", "541751": "Are the coupling constants in the dataset empirical coupling constants measured by NMR, or are they the results of some theoretical calculation with multiple (zero Kelvin, no solvent, DFT, etc. etc.) assumptions?", "648655": "I have little expertise in this domain, but I wonder what's the point of predicting scalar couplings given the 3-D structure of a molecule. As I know we can use NMR to gain insight into a molecule’s structure with the help of scalar couplings. But if we don't know the 3-D structure, how can we predict the scalar couplings. On the contrary, if we have known the 3D structure, there is no need to do this prediction. So I don't know what's the meaning of this competition task.", "561985": "I tried to run each cell of this kernel, but the cell related to \"folds\" is taking too much time(more than half an hour). Now I just commit the kernel but the processing is still going on. Could you please tell me how to figure out this problem.", "539443": "the level of theory should be provided for the potential energy calculations.", "560850": "", "554094": "", "560330": "Thanks!", "554217": "Thanks !", "553522": "Many thanks!", "547627": "Thanks." } }