A pyTorch Extension for Applied Mathematics
This version is compatible with pytorch (1.0.1) and later. You can create a conda environment for pytorch1:
conda create -n torch1 python=3 jupyter
source activate torch1
conda install pytorch=1 torchvision cudatoolkit=9.2 -c pytorch
# or conda install pytorch-cpu=1 -c pytorch
- aTEAM.optim.NumpyFuntionInterface: This function enable us to optimize pytorch modules with external optimizer such as scipy.optimize.lbfgsb.fmin_l_bfgs_b, see test/optim_quickstart.py
- aTEAM.nn.modules.MK: Moment matrix & convolution kernel convertor: aTEAM.nn.modules.MK.M2K, aTEAM.nn.module.MK.K2M
- aTEAM.nn.modules.Interpolation: Lagrange interpolation in a n-dimensional box: aTEAM.nn.modules.Interpolation.LagrangeInterp, aTEAM.nn.modules.Interpolation.LagrangeInterpFixInputs
- aTEAM.nn.functional.utils.tensordot: It is similar to numpy.tensordot
For more usages pls refer to aTEAM/test/*.py
aTEAM is a basic library for PDE-Net & PDE-Net 2.0(source code):
- PDE-Net: Learning PDEs from Data(ICML 2018)
Long Zichao, Lu Yiping, Ma Xianzhong, Dong Bin - PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network
Long Zichao, Lu Yiping, Dong Bin
If you find this code useful for your research then please cite
@inproceedings{long2018pdeI,
title={PDE-Net: Learning PDEs from Data},
author={Long, Zichao and Lu, Yiping and Ma, Xianzhong and Dong, Bin},
booktitle={International Conference on Machine Learning},
pages={3214--3222},
year={2018}
}
@article{long2018pdeII,
title={PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network},
author={Long, Zichao and Lu, Yiping and Dong, Bin},
journal={arXiv preprint arXiv:1812.04426},
year={2018}
}