hello-arun / ml-kirigami-tf

ML_for_kirigami_design (TensorFlow version)

Python package to model and to perform topology optimization for graphene kirigami using deep learning. We use convolutional neural networks (similar to VGGNet architecure) for regression.

General usage

• Python script to create and test fine grid dataset from coarse grid dataset is avalaible in bin\create-and-test-fine-grid.py

1. Regression and optimization

• A python code to perform regression with TensorFlow is avalaible in notebooks\CNN\01-regression.ipynb
• A python code to perform search optimal design with TensorFlow is avalaible in notebooks\CNN\02-search-designs.ipynb

2. Dataset

• Raw dataset of coarse-grained configs is avaialble in raw. This dataset generated using AIREBO potential with 1.7 mincutoff which is the default of CH.airebo.

This package is still under developement. More features will be added soon.

Building the conda environment

All dependencies are included in environment.yml and requirements.txt file. The enviroment can be generated by using

mamba env create --prefix ./env --file environment.yml --force

or

conda env create --prefix ./env --file environment.yml --force

If mamba do not exists then install it via

conda activate base
conda install mamba -n base -c conda-forge

Update the conda env

If you add (remove) dependencies to (from) the environment.yml file or the requirements.txt file after the environment has already been created, then you can re-create the environment with the following command.

conda activate base
mamba env update --prefix ./env --file environment.yml  --prune

References

• [1] Paul Z. Hanakata, E. D. Cubuk, D. K. Campbell, H.S. Park, Phys. Rev. Research, 2, 042006(R) (2020).
• [2] Paul Z. Hanakata, E. D. Cubuk, D. K. Campbell, H.S. Park, Phys. Rev. Lett, 121, 255304 (2018).
• [3] Paul Z. Hanakata