Fix typos in the rfm READMEs, add README to the fem and fix some typos.

Code for the publication 

Hiemer, Stefan, et al. "Predicting Creep Failure by Machine Learning--Which Features Matter?." arXiv preprint arXiv:2208.06923 (2022).

rfm contains the C++ codes to run the extremal simulations and creep

simulations as well as the machine learning and postprocessing scripts in

python. fem contains the ML and postprocessing scripts in python. 

 No newline at end of file

Get the data from the Zenodo: https://zenodo.org/record/7131771#.YzghRbTP1PY. 

To copy the paper, run the python scripts in this sequence:

- extract_data.py

- train_model.py

- save-predictions.py

- apply_model.py

- save-variances.py

Run then the individual figures as you please.

if __name__ == '__main__':

    #---- load simulation -----

    for ind in [2,1]:

    # create feature set 1

    datasets = ["csv_k1", "csv_k2", "csv_k8", "csv_k16",

                "csv_stress0.7","csv_stress0.9"]

    for dataset in datasets: 

        row_damage_postproc(h5file="row-new.h5", dataset=dataset,

                            pattern='*.csv', test = False)

    # create feature set 2        

    for w in [2,4,8,32]: 

        Michael_features(h5file="michael-X-100-width-"+str(w)+".h5",

                     datasets = ["csv_k1",

- train_model.py

- save_baseline.py

- save_predictions.py (adapt to your parameters e. g. temperature, current)

- summarize.py ()

- summarize.py

- rank-correlations.py

Run then the individual figures as you please.

Running the simulations is in run.py and the statistical information needed for the creep simulations are calculated by collect_data.py. Figure S1 can be created by fig-s1.py, but it the statistical information needs to be calculated first for that.

If you want to make sure that it works (in the sense that the program does not fail/break down) on your system, execute test.py. If you want to retrace the paper, execute redo, but be aware that it might take weeks/months/year on a simple work station, as we used a high perofmrance cluster for that.  

If you want to make sure that it works (in the sense that the program does not fail/break down) on your system, execute test.py. If you want to retrace the paper, execute redo, but be aware that it might take

weeks/months/years on a simple work station, as we used a high perofmrance cluster for that.