AI4SurrogatesMesoScale merge requestshttps://code.ornl.gov/7ml/ai4surrogatesmesoscale/-/merge_requests2022-10-28T12:45:10Zhttps://code.ornl.gov/7ml/ai4surrogatesmesoscale/-/merge_requests/4ShuffleStratified splitting + uncertainty quantification added2022-10-28T12:45:10ZLupo Pasini, MassimilianoShuffleStratified splitting + uncertainty quantification addedShuffle stratified splitting to:
- ensure that all bins are equally represented in training and testing portion of dataset
- K-fold cross validation
Average calculated to improve robustness of accuracy and precision evaluation
Standard ...Shuffle stratified splitting to:
- ensure that all bins are equally represented in training and testing portion of dataset
- K-fold cross validation
Average calculated to improve robustness of accuracy and precision evaluation
Standard deviation calculated to provide uncertainty of predictive performance of machine learning modelsLupo Pasini, MassimilianoLupo Pasini, Massimilianohttps://code.ornl.gov/7ml/ai4surrogatesmesoscale/-/merge_requests/3Max proposed changes atomic features2022-10-25T16:20:34ZLupo Pasini, MassimilianoMax proposed changes atomic features- added histogram for mixing - weak - repulsive categories
- check consistency between of supervised Chi2 multi-variate feature selection with sequence of supervised F-stat univariate feature selection- added histogram for mixing - weak - repulsive categories
- check consistency between of supervised Chi2 multi-variate feature selection with sequence of supervised F-stat univariate feature selectionLupo Pasini, MassimilianoLupo Pasini, Massimilianohttps://code.ornl.gov/7ml/ai4surrogatesmesoscale/-/merge_requests/2New notebook for Hmix multi-temp data2021-07-27T02:59:19ZChoi, Jong YoulNew notebook for Hmix multi-temp dataThis is to include notebooks to process multi-temp Hmix data: `Compositions+effective_elemental_properties_Hmix_800C+1100C+1200C.xlsx`
Compared with the notebooks in `Hmix_800C`, this version contains the following:
* Training set selec...This is to include notebooks to process multi-temp Hmix data: `Compositions+effective_elemental_properties_Hmix_800C+1100C+1200C.xlsx`
Compared with the notebooks in `Hmix_800C`, this version contains the following:
* Training set selection based on non-linear and linear unary /binary (`check-elemental.ipynb`)
* Fix on train/validation error calculation (`train.ipynb`)Choi, Jong YoulChoi, Jong Youlhttps://code.ornl.gov/7ml/ai4surrogatesmesoscale/-/merge_requests/1fixes and suggestions2021-07-12T14:57:36ZLupo Pasini, Massimilianofixes and suggestionsChoi, Jong YoulChoi, Jong Youl