Learning symmetries in datasets

Veronica Sanz

We investigate how symmetries present in datasets affect the structure of the latent space learned by Variational Autoencoders (VAEs). By training VAEs on data originating from sim…