Project Overview
This project introduces Algorithm-Informed Neural Networks (AINNs), a novel class of AI models that integrate algorithmic logic into their design and execution. AINNs aim to reduce data dependency, improve explainability, and enhance debuggability by aligning neural architectures with interpretable algorithmic structures.
Research Goals
- Develop open-source tools for designing and training AINNs
- Advance algorithmic theory by reverse engineering trained AINNs
- Demonstrate AINNs on NP-hard problems and real-world tasks
- Promote explainable and trustworthy AI in safety-critical domains
Broader Impacts
AINNs have the potential to transform both AI research and domain sciences, including computer algorithms and complexity theory. By extracting new algorithmic insights from learned models, this project seeks to generate breakthroughs in computational theory while enabling more transparent and reliable AI systems.
Collaboration and Outreach
We welcome partnerships with researchers, industry, and government agencies interested in explainable AI, algorithm design, and responsible deployment of intelligent systems. Educational resources and project tools will be freely available to the research community.
Project Team
Principal Investigator: Shahriar Nirjon (University of North Carolina at Chapel Hill)
Co-Principal Investigator: Md Yusuf Sarwar Uddin (University of Missouri-Kansas City)
Publications
Publications will be listed here as they become available.
Code and Demos
Links to code repositories, live demos, and videos will be added when available.