The Modelhub database documentation lists all the fields in the different Modelhub tables and indicates which differences there were between the ATM model and the one described in the related paper.
However, there seem to be some differences between the documented ATM model and the actual one.
I found the following differences:
Datasets:
dataset_id is called id in ATM
Order of the columns do not match (Not sure if it maters)
Dataruns:
datarun_id is called id in ATM
Order of the columns do not match (Not sure if it maters)
Hyperpartitions:
hyperpartition_id is called id in ATM
categoricals is called categorical_hyperparameters_64 in ATM
tunables is called tunable_hyperparameters_64 in ATM
constants is called constant_hyperparameters_64 in ATM
Classifiers:
classifier_id is called id in ATM
hyperparameters_values is called hyperparameter_values_64 in ATM
cv_judgment_metric_stdev was not in the old CSVs (should it say "not in the paper"?)
Order of the columns do not match (Not sure if it maters)
The Modelhub database documentation lists all the fields in the different Modelhub tables and indicates which differences there were between the ATM model and the one described in the related paper.
However, there seem to be some differences between the documented ATM model and the actual one.
I found the following differences:
dataset_idis calledidin ATMdatarun_idis calledidin ATMhyperpartition_idis calledidin ATMcategoricalsis calledcategorical_hyperparameters_64in ATMtunablesis calledtunable_hyperparameters_64in ATMconstantsis calledconstant_hyperparameters_64in ATMclassifier_idis calledidin ATMhyperparameters_valuesis calledhyperparameter_values_64in ATMcv_judgment_metric_stdevwas not in the old CSVs (should it say "not in the paper"?)