ML for Feynman Periods
Predicting Feynman periods in φ⁴-theory with machine learning
Feynman periods are rational numbers encoding the perturbative expansion of quantum field theories. Computing them analytically is extremely hard — they grow doubly exponentially in the loop order.
This project applies machine learning to predict Feynman periods from graph-theoretic features of the underlying Feynman diagrams. We show that neural networks can capture non-trivial combinatorial structure in φ⁴-theory, providing fast approximations that complement symbolic methods.
This work was done with Dr. Paul-Hermann Balduf during my MMath at the University of Waterloo.
Links: Paper (JHEP) · Thesis