Dataset: Validating thermodynamic models of arc-magma differentiation and training neural networks for rapid thermodynamic property inference

Description: This dataset has been produced as part of the research study titled ”Validating thermodynamic models of arc-magma differentiation and training neural networks for rapid thermodynamic property inference”. It contains the data generated and analyzed during the study, including then training and testing data, produced with MAGEMin, the surrogate model parameters obtained during cross-validation, and the parameters of the most accurate surrogate model. The dataset is intended to support transparency, reproducibility, and publication of the corresponding research article, which is currently submitted to a peer-reviewed journal. Contents: "TrainedModel_256HD_3HL_16384batchsize_1000epochs_full.jld2": Parameters of the most accurate surrogate model, as well as the data used during training "HyperParameterStudy_*": These files contain the surrogate model parameters as well as the training and validation data splits for each fold of the cross-validation. They also contain the training and validation losses for each fold "SynGEOROC_training_data*": This file contains the MAGEMin output data used to train the neural network "SynGEOROC_testing_data*": This file contains the MAGEMin output data used to test the surrogate models Usage Notes: Please refer to the README file and Julia scripts provided in the related repository for details on how to load and visualize the data. Once the related article is published, its DOI will be added to this record.