On Unsupervised Reconstruction with Dressed Multilayered Variational Quantum Circuits

The advantages of unsupervised quantum machine learning are still under study and appear to be very promising. Trainable variational quantum circuits are one example of successful approaches to combining classic machine learning and quantum. However, there is no clear path toward a quantum advantage for different types of variational circuits. This paper furthers the research efforts in understanding the potential and applications of hybrid quantum circuits. We study different circuits and see how similar they perform in an unsupervised learning task in an autoencoder configuration over a large multimodal dataset.