Synergistic Learning with Multi-Task DeepONet for Efficient PDE Problem Solving

This repository contains the source code and resources associated with the manuscript:
Synergistic Learning with Multi-Task DeepONet for Efficient PDE Problem Solving.

Synergistic.Learning.with.Multi-Task.DeepONet.for.Efficient.PDE.mp4

Overview

The repository demonstrates the implementation of Multi-Task DeepONet for solving partial differential equations (PDEs) efficiently through synergistic learning. The codebase is divided into three segments based on the problem discussed in the manuscript.

Installation

Prerequisites

• Conda
• Python 3.8 or higher

Setting up the Environment

1. Clone the repository:

   git clone https://github.com/varunsingh88/MT-DeepONet.git  
   cd MT-DeepONet
   conda env create -n "environment name" -f environment.yaml
   conda activate "environment name"

Downloading data

All dataset used in this study are available here MT-DeepONet data

Download individual data for each problem and store in the corresponding 'Data' folders.

Usage

Individual problems contained in the folders can be run independently after activating the conda environment:

python main.py

Checkpoints saved for each problem can be used for evaluating the model's output.

Reference

If you use this repository, please cite the manuscript:

@article{KUMAR2025107113,
title = {Synergistic learning with multi-task DeepONet for efficient PDE problem solving},
journal = {Neural Networks},
volume = {184},
pages = {107113},
year = {2025},
issn = {0893-6080},
doi = {https://doi.org/10.1016/j.neunet.2024.107113},
url = {https://www.sciencedirect.com/science/article/pii/S0893608024010426},
author = {Varun Kumar and Somdatta Goswami and Katiana Kontolati and Michael D. Shields and George Em Karniadakis},
}