Seminars

Home Centers & Programs AI and Natural Sciences Seminars

ABSTRACT: We present AI-based methods for holographic inverse problems within the gauge/gravity duality. Starting from boundary entanglement entropies S(l), we reconstruct the bulk metric function f(z) in (d+1)-dimensional asymptotically AdS black-hole backgrounds. A neural ansatz for f(z) is trained with a loss function consisting of smoothness regularization and the UV boundary condition f(0)=1. On test cases with unknown dual geometries, the recovered metrics reproduce the input source entanglement entropies. We then study holographic superconductors: in the probe limit on a Reissner–Nordström background with a momentum-relaxing real scalar and a charged complex scalar, we parameterize interactions via a function M. Learning M with a single KAN neural network yields accurate phase boundaries. Finally, we apply a Transformer model to learn mappings from boundary observables to bulk data. Together, these AI approaches enable exploration of strongly coupled quantum theories via holography.