Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Simple PyTorch Implementation of Physics Informed Neural Network (PINN)
| Date | Stars |
|---|---|
| 2026-07-31 | 379 |
| 2026-08-01 | 380 |
| 2026-08-05 | 380 |
| 2026-08-06 | 380 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# Simple PyTorch Implementation of Physics Informed Neural Network (PINN)
This repository contains my simple and clear to understand implementation of the paper [1]. As an example, I solved the 1D heat partial differential equation.
### 3D plot of the solution to 1D Heat Equation.
All the steps are included in a single file: `solve_PDE_NN.ipynb` which depends on the `pytorch` and `numpy` packages.
If you find this implementation useful, then please cite this work:
```
@software{Doloi_PyTorch_Implementation_of,
author = {Doloi, Nandita},
license = {MIT},
title = {{PyTorch Implementation of Physics Informed Neural Network (PINN)}}
}
```
This implementation was used in my following
[paper](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4477245). Here is the citation:
```
@article{doloi4477245iterative,
title={Iterative Methods for Neural Networks to Approximate the Solution of Nonlinear Two-Phase Flow Through Porous Media},
author={Doloi, Nandita and Phirani, Jyoti},
journal={Available at SSRN 4477245},
abstract = "{Machine Learning models are often used to make inferences on data that originate from some physical phenomena. For such applications, we might want the inferences to follow the laws of physics. Physics-informed neural networks (PINN) provide us with a mechanism to incorporate physical constraints on the data in the form of differential equations. In this work, we study the use of PINNs in the field of geological reservoirs particularly in solving the Buckley Leverette equation for nonlinear two-phase flow through porous media. PINNs have a tough time solving this equation as it belongs to the class of hyperbolic partial differential equations, which contain discontinuities in the solution. The discontinuity in the solution gives rise to a highly non-convex loss landscape for PINN, and so any second-order quasi-newton optimization method would get stuck in a local minimum and fail to find a good solution unless there is a very good initialization. In this work, we explore two methods which are Graduated optimization and Region based bootstrapping to handle this discontinuity and approximate a better solution to the Buckley Leverette equation as compared to the original PINN optimization. We observed that the loss we were able to achieve were at-least 10 times lower than the baseline.}"
}
```
[paper](https://onepetro.org/SPERCSC/proceedings-abstract/22RCSC/1-22RCSC/D011S001R006/515732?redirectedFrom=PDF). Here is the citation:
```
@proceedings{10.2118/212611-MS,
author = {Doloi, Nandita and Ghosh, Somnath and Phirani, Jyoti},
title = "{Super-Resolution Reconstruction of Reservoir Saturation Map with Physical Constraints Using Generative Adversarial Network}",
volume = {Day 1 Tue, January 24, 2023},
series = {SPE Reservoir Characterisation and Simulation Conference and Exhibition},
pages = {D011S001R006},
year = {2023},
month = {01},
abstract = "{Complete physics-based numerical simulations currently provide the most accurate approach for predicting fluid flow behavior in geological reservoirs. However, the amount of computational resources required to perform these simulations increase exponentially with the increase in resolution to the point that they are infeasible. Therefore, a common practice is to upscale the reservoir model to reduce the resolution such that numerous simulations, as required, can be performed within a reasonable time. The problem we are trying to solve here is that the simulation results from these upscaled models, although they provide a zoomed-out and global view of the reservoir dynamics, however, they lack a detailed zoomed-in view of a local region in the reservoir, which is required to take actionable decisions. This work proposes using super-resolution techniques, recently developed using machine learning methods, to obtain fine-scale flow behavior given flow behavior from a low-resolution simulExcerpt of 4,981 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:8be16bc172348463, topic:pytorch