Machine learning string standard models

We study machine learning of phenomenologically relevant properties string compactifications, which arise in the context heterotic line bundle models. Both supervised and unsupervised are considered. find that, for a fixed compactification manifold, relatively small neural networks capable distinguishing consistent models with correct gauge group chiral asymmetry from random without these properties. The same distinction can also be achieved learning, using an autoencoder. Learning nontopological properties, specifically number Higgs multiplets, turns out to more difficult, but is possible sizeable feature-enhanced datasets.