Mutual Information Estimation in Quantum Many-Body Systems using Transformers

This project was conducted as a research internship at Forschungszentrum Jülich’s PGI-8 Institute under the supervision of Dr. Markus Schmitt, from June 2024 to September 2024.

Project Overview

I investigated mutual information as a proxy to predict the difficulty of learning wave functions using Neural Quantum States (NQS) in quantum many-body systems using transformers. This research explored the intersection of quantum physics, information theory, and modern machine learning architectures.

Here’s the GitHub repository for this project: Neural Quantum States with Transformers

Key Contributions

• Designed and implemented transformer architectures in Julia to estimate entropy and conditional entropy in discrete and continuous quantum systems
• Collaborated with research team to enhance understanding of neural network representations of many-body quantum systems
• Applied scientific machine learning, information theory, and advanced statistical techniques to quantum systems

Technical Implementation

The project involved extensive work with:

• Julia programming for high-performance scientific computing
• Transformer networks adapted for quantum state representation
• Information theory metrics including mutual information, entropy, and conditional entropy
• Quantum many-body systems simulation and analysis
• Neural Quantum States (NQS) techniques to model quantum wave functions

Skills Developed

• Advanced Julia programming for scientific computing
• Deep understanding of transformer architectures in physics contexts
• Information theory applications in quantum mechanics
• Collaborative research in international academic environment