finalize

ToshiyukiBandai · Feb 10, 2022 · 5de1644 · 5de1644
1 parent f15072d
commit 5de1644
Show file tree

Hide file tree

Showing 29 changed files with 24,116 additions and 6,200 deletions.
diff --git a/PINNs_codes/.ipynb_checkpoints/inverse_heterogeneous-checkpoint.ipynb b/PINNs_codes/.ipynb_checkpoints/inverse_heterogeneous-checkpoint.ipynb
diff --git a/PINNs_codes/inverse_heterogeneous.ipynb b/PINNs_codes/inverse_heterogeneous.ipynb
diff --git a/README.md b/README.md
@@ -1,6 +1,6 @@
 # DD-PINNs-RRE: Physics-informed neural networks with Domain Decomposition for solving the Richardson-Richards equation
 
-This repository contains Python source codes for physics-informed neural networks (PINNs) with domain decomposition (DD) for solving the Richardson-Richards equation (RRE), which can be used to simulate water flow in soils. Also, this repository contains the analytical solutions to verify the PINN solutions as well as finite difference (written in Matlab) and finite element methods (using HYDRUS-1D) for comparison. The results are summarized in a paper that is currently reviewed by xxx.
+This repository contains Python source codes for physics-informed neural networks (PINNs) with domain decomposition (DD) for solving the Richardson-Richards equation (RRE), which can be used to simulate water flow in soils. Also, this repository contains the analytical solutions to verify the PINN solutions as well as finite difference (written in Matlab) and finite element methods (using HYDRUS-1D) for comparison. The results are summarized in a paper that is currently under reviewed.
 
 ### Contents
 * [PINNs_codes](#PINNs_codes)
@@ -46,7 +46,7 @@ year = {1991}
 }
 ```
 
-This directory contains numerical solutions by  [HYDRUS-1D](https://www.pc-progress.com/en/Default.aspx?hydrus-1d) used for inverse modeling to estimate surface water flux (true solution and noisy data). The HYDRUS output files are located in "results/inverse/HYDRUS."
+This directory contains numerical solutions by  [HYDRUS-1D](https://www.pc-progress.com/en/Default.aspx?hydrus-1d) used for inverse modeling to estimate surface water flux (true solution and noisy data). The HYDRUS output files are located in "results/inverse/HYDRUS." Note that the performance of PINNs is strongly affected by the initialization of neural networks, as dicsused in the paper. If you cannot obtain good results, please try different random seeds for the initialization.
 
 ## results
 This directory contains results shown in the paper.

diff --git a/results/example-inverse/1/K_pred.npy b/results/example-inverse/1/K_pred.npy
diff --git a/results/example-inverse/1/flux.png b/results/example-inverse/1/flux.png
diff --git a/results/example-inverse/1/flux_pred.npy b/results/example-inverse/1/flux_pred.npy
diff --git a/results/example-inverse/1/loss_data_1.npy b/results/example-inverse/1/loss_data_1.npy
diff --git a/results/example-inverse/1/loss_data_1_BFGS.npy b/results/example-inverse/1/loss_data_1_BFGS.npy
diff --git a/results/example-inverse/1/loss_data_2.npy b/results/example-inverse/1/loss_data_2.npy
diff --git a/results/example-inverse/1/loss_data_2_BFGS.npy b/results/example-inverse/1/loss_data_2_BFGS.npy
diff --git a/results/example-inverse/1/loss_psi_int.npy b/results/example-inverse/1/loss_psi_int.npy
diff --git a/results/example-inverse/1/loss_psi_int_BFGS.npy b/results/example-inverse/1/loss_psi_int_BFGS.npy
diff --git a/results/example-inverse/1/loss_qz_int.npy b/results/example-inverse/1/loss_qz_int.npy
diff --git a/results/example-inverse/1/loss_qz_int_BFGS.npy b/results/example-inverse/1/loss_qz_int_BFGS.npy
diff --git a/results/example-inverse/1/loss_res_1.npy b/results/example-inverse/1/loss_res_1.npy
diff --git a/results/example-inverse/1/loss_res_1_BFGS.npy b/results/example-inverse/1/loss_res_1_BFGS.npy
diff --git a/results/example-inverse/1/loss_res_2.npy b/results/example-inverse/1/loss_res_2.npy
diff --git a/results/example-inverse/1/loss_res_2_BFGS.npy b/results/example-inverse/1/loss_res_2_BFGS.npy
diff --git a/results/example-inverse/1/loss_res_int.npy b/results/example-inverse/1/loss_res_int.npy
diff --git a/results/example-inverse/1/loss_res_int_BFGS.npy b/results/example-inverse/1/loss_res_int_BFGS.npy
diff --git a/results/example-inverse/1/psi_color.png b/results/example-inverse/1/psi_color.png
diff --git a/results/example-inverse/1/psi_pred.npy b/results/example-inverse/1/psi_pred.npy
diff --git a/results/example-inverse/1/residual.npy b/results/example-inverse/1/residual.npy
diff --git a/results/example-inverse/1/residual.png b/results/example-inverse/1/residual.png
diff --git a/results/example-inverse/1/theta_color.png b/results/example-inverse/1/theta_color.png
diff --git a/results/example-inverse/1/theta_fixed_time.png b/results/example-inverse/1/theta_fixed_time.png
diff --git a/results/example-inverse/1/theta_pred.npy b/results/example-inverse/1/theta_pred.npy
diff --git a/results/example-inverse/1/training.png b/results/example-inverse/1/training.png
diff --git a/results/example-inverse/summary.xls b/results/example-inverse/summary.xls