Predict_3cation_testtrain_69ternaries.csv

The file contains an index column followed by 39 columns name like <element symbol.AtfRac, e.g. "Ag.AtFrac". These are normalized cation concentrations.

There are 69 data instances such as "Ag-Bi-Mn" where the values are:
"test" means this is a composition containing Ag,Bi,Mn that is on a 10 atom% composition grid in the 3-cation space
"train" means the this composition is available for training and/or validation
"NA" means for this data instance this composition shouldn't be used because it contans Ag, Bi, and Mn. It may contain other elements too.

Finally there are 10 columns alpha__<index> that give the experimental absorption spectrum at these 10 photo energies:
1.39, 1.58, 1.77, 1.96, 2.15, 2.35, 2.54, 2.73, 2.92, 3.11

==========================================
results_69ternaries.csv

Has prediction results for all 69 data instances and many models. Each (composition,alpha index) is a row with ternary_data_instance matching the column from the above file where the "test" composition came from.
The standardized_ground_truth transforms the value using the data instance and alpha index-specific mean_standardization and std_standardization. All prediction results are in these standardized units.

==========================================
results_random_split.csv

Similar results files but with a Random test set chosen from the union of the 69 data instance test sets.

=========================================
prediction_new_ternary_compositions_hclmpt.csv

Of the 31 elementws in the 69 data instances, 31-choose-3 minus 69 2-cation instances with 10% grid are predicted using the H-CLMP(T) (the same thing as HCLMP-GenDOS) model using the entire avaialble dataset as training data.