I am a data science professional with a PhD in Instrumentation (machine learning applied to pharmacology) and previous experience in applying evolutionary algorithms to empirical and theoretical chemistry problems (coherent control and finding saddle points on PES). \nI am looking to partner with someone who can derive useful features and provide insight from the physical chemistry side. For example, \n How can we make use of the Mulliken charges provided for the training set? For example, one kernel calculates grouped statistics by atom type and maps these to the test data, can we improve on this?\n How can we make use of the scalar coupling constant contributions data? For example, would it make sense to estimate a multiheaded neural network that estimates the contributions on the train set and the sum on the train and test sets?\n What QM can we do quickly on the train and structure data that we can also do on the test data and would be useful for predicting scalar coupling constant?\nPlease email me if interested: robert.burbidge@yahoo.co.uk.
", "messages": [ { "id": "541357", "postDate": "06/02/2019 09:43:30", "content": "I am a data science professional with a PhD in Instrumentation (machine learning applied to pharmacology) and previous experience in applying evolutionary algorithms to empirical and theoretical chemistry problems (coherent control and finding saddle points on PES). \nI am looking to partner with someone who can derive useful features and provide insight from the physical chemistry side. For example, \n How can we make use of the Mulliken charges provided for the training set? For example, one kernel calculates grouped statistics by atom type and maps these to the test data, can we improve on this?\n How can we make use of the scalar coupling constant contributions data? For example, would it make sense to estimate a multiheaded neural network that estimates the contributions on the train set and the sum on the train and test sets?\n What QM can we do quickly on the train and structure data that we can also do on the test data and would be useful for predicting scalar coupling constant?\nPlease email me if interested: robert.burbidge@yahoo.co.uk.
", "rawMarkdown": "I am a data science professional with a PhD in Instrumentation (machine learning applied to pharmacology) and previous experience in applying evolutionary algorithms to empirical and theoretical chemistry problems (coherent control and finding saddle points on PES). \nI am looking to partner with someone who can derive useful features and provide insight from the physical chemistry side. For example, \n How can we make use of the Mulliken charges provided for the training set? For example, one kernel calculates grouped statistics by atom type and maps these to the test data, can we improve on this?\n How can we make use of the scalar coupling constant contributions data? For example, would it make sense to estimate a multiheaded neural network that estimates the contributions on the train set and the sum on the train and test sets?\n What QM can we do quickly on the train and structure data that we can also do on the test data and would be useful for predicting scalar coupling constant?\nPlease email me if interested: robert.burbidge@yahoo.co.uk.", "votes": null }, { "id": "541486", "postDate": "06/02/2019 14:24:06", "content": "I'm a PhD physicist who has 14 years of NMR experience, my thesis was on homonuclear correlations (e.g. 2JHH). I know some python/pandas/matplotlib but haven't performed any machine learning.\nDo you know much about graph CNN?\nAlso, we can't do any QM... Quoting the rules: \"You may use data other than the Competition Data (“External Data”) to develop and test your models and Submissions excluding data from quantum mechanics computations made specifically for the competition\"
", "rawMarkdown": "I'm a PhD physicist who has 14 years of NMR experience, my thesis was on homonuclear correlations (e.g. 2JHH). I know some python/pandas/matplotlib but haven't performed any machine learning.\nDo you know much about graph CNN?\nAlso, we can't do any QM... Quoting the rules: \"You may use data other than the Competition Data (“External Data”) to develop and test your models and Submissions excluding data from quantum mechanics computations made specifically for the competition\"", "votes": null }, { "id": "541505", "postDate": "06/02/2019 15:05:48", "content": "Good point. I was being a little loose in my terminology. I meant more things like using the Karplus equation than heavy computational chemistry with, say, GAUSSIAN.
", "rawMarkdown": "Good point. I was being a little loose in my terminology. I meant more things like using the Karplus equation than heavy computational chemistry with, say, GAUSSIAN.", "votes": null } ], "comments": [ { "id": 541486, "author_name": "phucco", "author_url": "", "post_date": "06/02/2019 14:24:06", "content": "I'm a PhD physicist who has 14 years of NMR experience, my thesis was on homonuclear correlations (e.g. 2JHH). I know some python/pandas/matplotlib but haven't performed any machine learning.\nDo you know much about graph CNN?\nAlso, we can't do any QM... Quoting the rules: \"You may use data other than the Competition Data (“External Data”) to develop and test your models and Submissions excluding data from quantum mechanics computations made specifically for the competition\"
", "votes": null, "replies": [] }, { "id": 541505, "author_name": "robertburbidge", "author_url": "", "post_date": "06/02/2019 15:05:48", "content": "Good point. I was being a little loose in my terminology. I meant more things like using the Karplus equation than heavy computational chemistry with, say, GAUSSIAN.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "541357": "I am a data science professional with a PhD in Instrumentation (machine learning applied to pharmacology) and previous experience in applying evolutionary algorithms to empirical and theoretical chemistry problems (coherent control and finding saddle points on PES). \nI am looking to partner with someone who can derive useful features and provide insight from the physical chemistry side. For example, \n How can we make use of the Mulliken charges provided for the training set? For example, one kernel calculates grouped statistics by atom type and maps these to the test data, can we improve on this?\n How can we make use of the scalar coupling constant contributions data? For example, would it make sense to estimate a multiheaded neural network that estimates the contributions on the train set and the sum on the train and test sets?\n What QM can we do quickly on the train and structure data that we can also do on the test data and would be useful for predicting scalar coupling constant?\nPlease email me if interested: robert.burbidge@yahoo.co.uk.", "541486": "I'm a PhD physicist who has 14 years of NMR experience, my thesis was on homonuclear correlations (e.g. 2JHH). I know some python/pandas/matplotlib but haven't performed any machine learning.\nDo you know much about graph CNN?\nAlso, we can't do any QM... Quoting the rules: \"You may use data other than the Competition Data (“External Data”) to develop and test your models and Submissions excluding data from quantum mechanics computations made specifically for the competition\"", "541505": "Good point. I was being a little loose in my terminology. I meant more things like using the Karplus equation than heavy computational chemistry with, say, GAUSSIAN." }, "source": "meta" }