Symbolic artificial intelligence for hidden topological orders in quantum physics

Artificial intelligence (AI) offers novel methodologies to unravel complex physical phenomena. However, most machine learning models lack transparency in their decision-making processes. In this proposal, we aim to develop a symbolic AI as a tool to reveal hidden topological orders in quantum physics.

To this end, an AI-assisted symbolic regression method will be studied.

We focus on three main objectives: (i) machine learning topological phases with experimental data; (ii) uncovering hidden non-local symmetry-protected topological orders; and (iii) searching for quantized topological invariants in an unsupervised fashion. The interplay between symbolic AI and quantum physics is envisioned to bring new insights into topological phases.

Moreover, the project will scrutinize the explainability and the robustness of machine learning models. The investigations will provide concrete guidelines for accompanying theoretical and experimental studies at MagTop.

The outcomes of the project will pave the way to discover novel features of topological materials in a reliable and explainable way, as well as provide great opportunities for me to reach a position of professional excellence and independence.