Mutual information maximizing quantum generative adversarial networks
This work addresses a specific problem in quantum machine learning for researchers in the NISQ era, offering an incremental improvement by combining existing InfoGAN principles with QGANs.
The authors tackled mode collapse and lack of feature control in quantum generative adversarial networks (QGANs) by proposing InfoQGAN, a hybrid quantum-classical model that integrates mutual information optimization, resulting in mitigated mode collapse and robust feature disentanglement in numerical simulations on synthetic 2D distributions and the Iris dataset.
One of the most promising applications in the era of Noisy Intermediate-Scale Quantum (NISQ) computing is quantum generative adversarial networks (QGANs), which offer significant quantum advantages over classical machine learning in various domains. However, QGANs suffer from mode collapse and lack explicit control over the features of generated outputs. To overcome these limitations, we propose InfoQGAN, a novel quantum-classical hybrid generative adversarial network that integrates the principles of InfoGAN with a QGAN architecture. Our approach employs a variational quantum circuit for data generation, a classical discriminator, and a Mutual Information Neural Estimator (MINE) to explicitly optimize the mutual information between latent codes and generated samples. Numerical simulations on synthetic 2D distributions and Iris dataset augmentation demonstrate that InfoQGAN effectively mitigates mode collapse while achieving robust feature disentanglement in the quantum generator. By leveraging these advantages, InfoQGAN not only enhances training stability but also improves data augmentation performance through controlled feature generation. These results highlight the potential of InfoQGAN as a foundational approach for advancing quantum generative modeling in the NISQ era.