Speaker
Description
Variational quantum circuits (VQCs) are among the most promising frameworks for near-term quantum computing, offering a flexible and hardware-efficient approach to leveraging current noisy intermediate-scale quantum (NISQ) devices. In this talk, we provide a structured overview of VQC architecture, discussing its key building blocks: the data encoding stage, which embeds classical information in the quantum Hilbert space, and the parameterized ansatz, which constitutes the trainable core of the circuit and is optimized via a classical feedback loop.
Building on this foundation, we present two practical applications of the VQC paradigm. The first is a hybrid quantum-classical convolutional neural network (QCNN), in which quantum convolutional and pooling layers are designed to extract features from structured data while progressively reducing the size of the quantum register. The second is a quantum generative adversarial network (QGAN), in which a quantum generator learns to produce data distributions that a classical or quantum discriminator cannot distinguish from real data.
For both architectures, we discuss simulation results that validate the proposed circuit designs and training strategies. We then present early experimental results obtained on IQM's EMERALD quantum processor, providing an initial assessment of how these models perform on real superconducting hardware. The comparison between simulated and hardware-executed runs provides insight into the impact of device noise on training dynamics and output quality, reinforcing the importance of beyond-simulation benchmarking for variational quantum machine learning.