5 The Phase Separation Binary Classifier: where to read more about it

There are quite a few places that you can read more about the Phase Separation Binary Classifier (PSBC), and also see further examples.

• I have posted a paper with all the mathematics behind the model. It is quite self contained. There is a preprint currently available on arXiv at https://arxiv.org/abs/2009.02467.

• All the data is available at (Monteiro 2020b).

• There are many other examples of usage in the jupyter-notebooks:

  • PSBC_Examples.ipynb is a short tutorial that explains how to use the model in python.
  • PSBC_grid_search_notebook.ipynb shows you how to replicate the grid search in order to find good hyperaparamters.
  • PSBC_ensemble_learning_notebook.ipynb explains how Ensemble learning was used in practice, besides showing how confusion matrices where created.
  • PSBC_training_notebook.ipynb can be used to train the model once grid search has been carried out. Actually, if you have the output of grid search already in hands, you can run this nootebook without runing th grid search notebook.
  • PSBC_using_statistical_files.ipynb shows how the data we anbalize in the paper ccan be retrieved from pickled files.

In order to keep things short I didn’t explain how callbacks are used in order to save validation data statistics. This was necessary because the keras/tensorflow API does not support this feature in “non-standard” models, like ours. If you want to see how this is done, please check both files tf_PSBC_extra_libs_for_training_and_grid_search.py and tfversion_binary_phase_separation.py at (Monteiro 2020a).

• If you want to have access to the trainable examples I used, and to the computational statistics, you can either download the files PSBC_BCs.tar.gz, about 11Mn for the tests regards different boundary conditions, and PSBC_classifier_PCA.tar.gz, about 2 Mb at the companion data repository to this project at Zenodo.
  
• In this Github you will also find a manual named README_v2.pdf to the data in the repository.

References

Monteiro, Rafael. 2020a. “Source Code for the Paper ‘Binary Classification as a Phase Separation Process’.” GitHub Repository. https://github.com/rafael-a-monteiro-math/Binary_classification_phase_separation; GitHub.

Monteiro, Rafael. 2020b. “Binary Classification as a Phase Separation Process (data repository).” Zenodo. https://doi.org/10.5281/zenodo.5525794.