Physically Interpretable Representation Learning with Gaussian Mixture Variational AutoEncoder (GM-VAE)