Three-body renormalization group limit cycles based on unsupervised feature learning

Abstract Both the three-body system and inverse square potential carry a special significance in study of renormalization group limit cycles. In this work, we pursue an exploratory approach address question which two-body interactions lead to cycles at low energies, without imposing any restrictions upon scattering length. For this, train boosted ensemble variational autoencoders, that not only provide severe dimensionality reduction, but also allow generate further synthetic potentials, is important prerequisite order efficiently search for low-dimensional latent space. We do so by applying elitist genetic algorithm population potentials minimizes specially defined limit-cycle-loss. The resulting fittest individuals suggest cycle loss independent hyperangle.