Hybrid Quantum Variational Autoencoders for Representation Learning