AI Quantum Entangled Pet
Why did you apply for this internship?
I applied to this internship to develop my BSc project into publishable scientific research, while acquiring invaluable computational skills to complement my physics background.
What did you hope to gain in completing this project?
By completing this project, I hoped to gain valuable experience in machine learning and academic writing, culminating in a research paper suitable for publication in a peer-reviewed journal.
Project Overview
This project builds on my BSc research, which identified a key challenge in training neural networks to predict gauge-dependent electronic properties of finite quantum systems within density functional theory, even when the gauge is fixed.
The project broadens this study to a wider range of quantum systems by modelling larger numbers of electrons, and assessing whether my findings generalise to other machine learning models (kernel rider regression). The focus is to strengthen my findings through rigorous validation so they are suitable for publication in a peer-reviewed journal.
What were the key results of your research project?
My finding that gauge-dependent density functionals are harder to model accurately with machine learning generalises to systems of multiple interacting electrons and to different machine learning models, indicating a pervasive problem. The root cause of this problem are specific transformations in the external potential ('near-constant shifts') that cause small changes in the electron density and disproportionately large changes to gauge-dependent properties. I formally defined near-constant shifts and established a clear link to machine learning performance.
GitHub repository: https://github.com/LeoArnstein/gauge-dependent-density-functionals
How do you feel you have benefited from completing this internship and has it made you consider future career paths?
This internship has provided me with the opportunity to strengthen the findings of my BSc research for submission to a peer-reviewed journal, which could open up further research opportunities and have a significant impact on my future. It has also enabled me to develop my skills in machine learning, applying models and techniques that I hadn't used previously.
Before the internship, I was not aware of the research software engineering role, and am now considering it as a potential career path.