Speaker
Haruto Kitagawa
(Kyushu University)
Description
In this study, we analyze the latent features of physical observables in the lepton sector using an autoencoder. To obtain the dataset of physical observables, we employ flow matching to globally sample the Yukawa matrices and Majorana mass parameters within the framework of the type-I seesaw mechanism, reproducing the experimentally observed values. The learned latent representation reveals previously unknown correlations among physical observables. These findings may provide new insights into the origin of the neutrino mass hierarchy and the lepton mixing pattern.