Level Set Learning with Pseudo-Reversible Neural Networks for Nonlinear Dimension Reduction in Function Approximation

Code repository for the paper:
Level Set Learning with Pseudo-Reversible Neural Networks for Nonlinear Dimension Reduction in Function Approximation
Yuankai Teng, Zhu Wang, Lili Ju, Anthony Gruber, Guannan Zhang
SIAM Journal on Scientific Computing, 2023
[paper]

Training Usage

To train the PRNN for a problem on given domain and draw a graph for regression

python ./train_model.py
 --case 2 
 --dim 2 
 --hidden_layers 2 
 --hidden_neurons 20 
 --lam_adf 1 
 --lam_bd 1 
 --optimizer 'Adam' 
 --Test_Mode 'LocalFitting' 
 --epochs_Adam 5000 
 --epochs_LBFGS 200 
 --TrainNum 2000 
 --coeff_para 50 
 --sigma 0.01 
 --domain 0 1

Testing Usage

To evaluate numerical error and relative sensitivity

python ./evaluate_model.py
 --case 2 
 --dim 2 
 --hidden_layers 2 
 --hidden_neurons 20 
 --lam_adf 1 
 --lam_bd 1 
 --optimizer 'Adam' 
 --Test_Mode 'LocalFitting' 
 --epochs_Adam 5000 
 --epochs_LBFGS 200 
 --TrainNum 2000 
 --coeff_para 50 
 --sigma 0.01 
 --domain 0 1

Citation

If you find the idea or code of this paper useful for your research, please consider citing us:

@article{teng2023level,
  title={Level Set Learning with Pseudoreversible Neural Networks for Nonlinear Dimension Reduction in Function Approximation},
  author={Teng, Yuankai and Wang, Zhu and Ju, Lili and Gruber, Anthony and Zhang, Guannan},
  journal={SIAM Journal on Scientific Computing},
  volume={45},
  number={3},
  pages={A1148--A1171},
  year={2023},
  publisher={SIAM}
}