DiMoGraph Active Learning of Gaussian Process Regression
Version: v0.1.0
Author(s): Florian Fuchs <florian.fuchs@enas.fraunhofer.de>, Fabian Teichert <fabian.teichert@enas.fraunhofer.de>, Philipp Schulze <pschulze@math.tu-berlin.de>
pyiron scikit-learn owlready2 Gaussian process regression active learning machine learning
Active learning of Gaussian Process Regression based on a network model using the workflow manager pyiron.
README.md
Gaussian process regression example for pyiron
This repository contains a Jupyter notebook workflow.ipynb which demonstrates a pyiron workflow of performing an adaptive training. The training data are created using a simplified network model, which calculates the electrical conductivity from a given network of conductors. The workflows are implemented using pyiron.
The corresponding pyiron classes are NetworkModelJob in code/pyiron_networkmodel/pyiron_networkmodel.py and AdaptiveGaussianProcessRegressionJob in code/pyiron_adaptive_gaussian_process_regression/adaptive_gaussian_process_regression.py.
Data from the training data creation as well as the adaptive training are written to an ontology.
This workflow has been developed within the BMFTR-funded project DiMoGraph (Digital models for graphene-based conductor materials, 13XP5190B).
Contributers:
Florian Fuchs (Fraunhofer ENAS, Chemnitz)
Fabian Teichert (Fraunhofer ENAS, Chemnitz)
Philipp Schulze (TU Berlin, Berlin)
Requirements:
pyiron
pyiron_base
scikit-learn
numpy
optuna
owlready2