Inverse-designed spinodoid metamaterials

Implementation of machine learning framework for "Inverse-designed Spinodoid Metamaterials" (https://doi.org/10.1038/s41524-020-0341-6) as described in the following publication.

Citation

If you use this code, please cite the following publication: S. Kumar, S. Tan, L. Zheng, D.M. Kochmann, Inverse-designed spinodoid metamaterials, npj Comput Mater 6, 73 (2020). https://doi.org/10.1038/s41524-020-0341-6

Note: To generate the spinodoid designs (predicted from machine learning), see: https://www.gibboncode.org/html/HELP_spinodoid.html.

Requirements

• Python (tested on version 3.7.1)
• Python packages:
  • PyTorch (tested without CUDA)
  • NumPy
  • pandas
  • statistics

Usage

python main.py

File descriptions

• main.py: main file to be executed and contains training protocols
• model.py: functions for creating neural network models
• loadDataset.py: functions for loading data from data.csv
• errorAnalysis.py: functions for post-processing and error analysis
• normalization.py: functions for normalization of features (inputs to neural networks)
• parameters.py: contains all parameters and hyper-parameters for neural network architectures and training protocols

Outputs

After training is over, outputs will be available in the following directories:

• ./models/ : contains trained models
• ./loss-history/ : contains loss history during training