[HEG] Higher dimensional conformal fields from neural network statistics and applications

Who: Joydeep Naskar (Northeastern U.)

Title: Higher dimensional conformal fields from neural network statistics and applications

Abstract: In this talk, I will discuss the motivations and framework of neural network-field theories. I will construct simple examples of non-unitary conformal fields and compute their two-, three- and four- point correlators in 4-dimensions. I will discuss the operator spectrum of the constructed example using conformal block decomposition and show a precise agreement with a notion of fusion rules. Next, I will give the example of a free boson and move to spinning fields. Finally, I will return to the motivation of using neural-networks for field theory applications, by giving an explicit example of a trained neural network approximating the solution to the Navier-Stokes in 3 spatial dimensions, matching the scaling laws predicted by Kolmogorov.