Events

Date: 24 April 2026   Time: 14:30 - 15:30

Location: 114 G.O. Jones Building

Title: Neural Networks and Quantum Field Theory

Abstract: The neural network-field theory (NN-FT) correspondence proposes that certain quantum field theories can be defined by specifying an appropriate neural network architecture and parameter density. I will review some of the recent progress in this area, including the construction of the 2d Liouville conformal field theory. Interactions in Liouville theory can be implemented by breaking statistical independence of neural network parameters in a prescribed way, allowing numerical computation of the 3-point function of vertex operators on the sphere which reproduces the DOZZ formula. Topological effects in field theories, such as the vortices involved in the Berezinskii-Kosterlitz-Thouless transition and the T-duality of the bosonic string, can be incorporated into the NN-FT paradigm by including additional discrete parameters that label topological sectors. Based on 2601.14453 and 2604.02313.