PyTorch Tutorial for Physics-informed Machine Learning

This repository contains several key models in Physice-informed machine learning (PI-ML) and data-driven machine learning written in PyTorch.

Data-driven models

Neural Network (NN)
- Standard fully connected neural networks
Multi-fidelity Neural Network (MFNN)
- Three standard neural networks coupled to fit high-fidelity data, high-fidelity data and their linear combination.
Convolutional Neural Network (CNN)
- Convolutional neural network(Decoder)

Physics-informed Machine Learning (PIML)

Physical-informed Neural Networks (PINNs)
- Physical-informed neural network for solving partial differential equations, e.g., Allen-Cahn equation(1D time-dependent and 2D equilibrium state)
Deep Operator Networks (DeepONet)
- DeepONet for learning a PDE operator

Proof of Concept are listed below:

DynNet (Dynamic-graph Network)
- Fully-connected neural network to demonstrate the concept of dynamic graph.
Gradient (Automatic Differentiation)
- Calculate gradient in PyTorch