Hi All, I am curious that why we only one potential, dipole moment and mulliken charges... information for train dataset? Is that possible to get all of the information for test dataset?
", "messages": [ { "id": "583793", "postDate": "07/25/2019 01:48:20", "content": "Hi All, I am curious that why we only one potential, dipole moment and mulliken charges... information for train dataset? Is that possible to get all of the information for test dataset?
", "rawMarkdown": "Hi All, I am curious that why we only one potential, dipole moment and mulliken charges... information for train dataset? Is that possible to get all of the information for test dataset?", "votes": null }, { "id": "583819", "postDate": "07/25/2019 03:18:36", "content": "https://www.kaggle.com/artgor/using-meta-features-to-improve-model
", "rawMarkdown": "https://www.kaggle.com/artgor/using-meta-features-to-improve-model", "votes": null }, { "id": "583898", "postDate": "07/25/2019 06:26:07", "content": "You can also extract the Mulliken charges and dipole moment from QM9 (although be careful about the direction of the DM: https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836)
", "rawMarkdown": "You can also extract the Mulliken charges and dipole moment from QM9 (although be careful about the direction of the DM: https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836)", "votes": null }, { "id": "584106", "postDate": "07/25/2019 12:43:39", "content": "The the first line after Additional Data title in the Data section -\nNOTE: additional data is provided for the molecules in Train only!
", "rawMarkdown": "The the first line after Additional Data title in the Data section -\nNOTE: additional data is provided for the molecules in Train only!", "votes": null }, { "id": "584123", "postDate": "07/25/2019 13:15:22", "content": "Hmm, however it took some time to compute QM9 so if these features are used in a model without the need to predict it by another model in the pipeline, then I do not see the point of the competition. Every such model then need to do QM9-level computations to obtain these features. ML then do not spare time.
", "rawMarkdown": "Hmm, however it took some time to compute QM9 so if these features are used in a model without the need to predict it by another model in the pipeline, then I do not see the point of the competition. Every such model then need to do QM9-level computations to obtain these features. ML then do not spare time.", "votes": null }, { "id": "584167", "postDate": "07/25/2019 14:10:22", "content": "Maybe. On the other hand, nothing in QM9 seems to really be very useful for calculating scalar-coupling, so it's probably a moot point.
", "rawMarkdown": "Maybe. On the other hand, nothing in QM9 seems to really be very useful for calculating scalar-coupling, so it's probably a moot point.", "votes": null } ], "comments": [ { "id": 583819, "author_name": "artgor", "author_url": "", "post_date": "07/25/2019 03:18:36", "content": "https://www.kaggle.com/artgor/using-meta-features-to-improve-model
", "votes": null, "replies": [] }, { "id": 583898, "author_name": "fnands", "author_url": "", "post_date": "07/25/2019 06:26:07", "content": "You can also extract the Mulliken charges and dipole moment from QM9 (although be careful about the direction of the DM: https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836)
", "votes": null, "replies": [ { "id": 584123, "author_name": "somnambwl", "author_url": "", "post_date": "07/25/2019 13:15:22", "content": "Hmm, however it took some time to compute QM9 so if these features are used in a model without the need to predict it by another model in the pipeline, then I do not see the point of the competition. Every such model then need to do QM9-level computations to obtain these features. ML then do not spare time.
", "votes": null, "replies": [] }, { "id": 584167, "author_name": "fnands", "author_url": "", "post_date": "07/25/2019 14:10:22", "content": "Maybe. On the other hand, nothing in QM9 seems to really be very useful for calculating scalar-coupling, so it's probably a moot point.
", "votes": null, "replies": [] } ] }, { "id": 584106, "author_name": "scirpus", "author_url": "", "post_date": "07/25/2019 12:43:39", "content": "The the first line after Additional Data title in the Data section -\nNOTE: additional data is provided for the molecules in Train only!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "583793": "Hi All, I am curious that why we only one potential, dipole moment and mulliken charges... information for train dataset? Is that possible to get all of the information for test dataset?", "583819": "https://www.kaggle.com/artgor/using-meta-features-to-improve-model", "583898": "You can also extract the Mulliken charges and dipole moment from QM9 (although be careful about the direction of the DM: https://www.kaggle.com/c/champs-scalar-coupling/discussion/99055#latest-572836)", "584106": "The the first line after Additional Data title in the Data section -\nNOTE: additional data is provided for the molecules in Train only!", "584123": "Hmm, however it took some time to compute QM9 so if these features are used in a model without the need to predict it by another model in the pipeline, then I do not see the point of the competition. Every such model then need to do QM9-level computations to obtain these features. ML then do not spare time.", "584167": "Maybe. On the other hand, nothing in QM9 seems to really be very useful for calculating scalar-coupling, so it's probably a moot point." }, "source": "meta" }