Guided Variational Autoencoder for Disentanglement Learning