Speaker
Description
In strongly coupled field theories, perturbation theory cannot be employed to study the low-energy spectrum. Thus, non-perturbative techniques are required. One possibility is the Lagrangian approach, where energies are extracted from the Euclidean-time dependence of correlation functions. This method suffers from excited-state contamination at shorter times and rapidly growing statistical noise at larger times, leaving only a narrow time window to extract the energy of the system. An alternative is the variational approach in the Hamiltonian formalism, which does not present such signal-to-noise problem. However, it recquires the choice of a trial wave function. In this work, we study the viability of employing a neural network as a variational ansatz. As a first step towards more phenomenologically interesting strongly coupled theories, like quantum chromodynamics, we study scalar field theories with quartic couplings.
