HOME > MODNet

MODNet

MODNet (Material Optimal Descriptor Network) is a supervised machine-learning framework for predicting materials properties from either composition or crystal structure. It combines matminer-based feature generation with mutual-information feature selection and joint learning to deliver strong results on limited datasets.

Information
Case / Article
Learning Content
Review
Developer’s Voice

Information

Official site	https://github.com/ppdebreuck/modnet
Openness	★★★
License	MIT License
Core Developers	Developers: Pierre-Paul De Breuck and Matthew Evans, with contributions from David Waroquiers, Gregoire Heymans, and the MODNet community.
Related Papers	Pierre-Paul De Breuck et al., “Materials property prediction for limited datasets enabled by feature selection and joint learning with MODNet,” npj Computational Materials 7, 83 (2021). doi:10.1038/s41524-021-00552-2 Pierre-Paul De Breuck et al., “Robust model benchmarking and bias-imbalance in data-driven materials science: a case study on MODNet,” Journal of Physics: Condensed Matter 33, 404002 (2021). doi:10.1088/1361-648X/ac1280
Other	Description: MODNet packages a full workflow for materials property prediction. `MODData` handles featurization (composition or structure) using matminer descriptors, then performs feature selection to produce a compact, information-rich descriptor set. `MODNetModel` trains a neural network that can support single or multiple target properties, making joint learning practical for small or noisy datasets. The repository also ships pretrained models for refractive index and vibrational thermodynamics, and integrates with MatBench benchmarks.