Hybrid Classical-Quantum Variational Autoencoder for Neural Topic Modeling

The article presents a hybrid classical-quantum variational autoencoder (VAE) for neural topic modeling, integrating parameterized quantum circuits within the VAE's inference network while maintaining a classical topic-word decoder. It introduces a modified Gaussian Softmax posterior that allows operation on a 10-qubit quantum device, achieving superior performance on the AgNews dataset with a $C_v$ coherence score of 0.71 and an NPMI score of 0.20 compared to state-of-the-art neural topic models. This work highlights the potential of hybrid VAEs for effective topic modeling on near-term quantum hardware, suggesting a viable path for quantum-enhanced machine learning applications.