Variational Quantum Algorithms (VQAs) are a promising approach to leverage Noisy Intermediate-Scale Quantum computers. But choosing optimal quantum circuits that efficiently solve a given VQA problem is a non-trivial task. However, Quantum Architecture Search (QAS) algorithms enable automatic generation of quantum circuits tailored to the provided problem. Existing QAS approaches typically adapt classical neural network architecture search techniques to train machine learning models to sample relevant circuits, often overlooking the inherent quantum nature of QAS. By embracing the quantum nature of the search space directly, we propose a sampling-free differentiable QAS algorithm that models the search process as the evolution of a mixed state, which naturally emerges from the search space of quantum circuits. Further, the mixed state formulation enables our method to incorporate generic quantum noise channels. Using numerical simulations, we validate our method by generating circuits for state initialization and Hamiltonian optimization tasks. We show that our approach is comparable to, if not outperforming, existing QAS techniques, shows improved robustness to noise, and requires significantly fewer quantum simulations during training.
@article{kumar2025rhodartsdifferentiablequantumarchitecture,
title={RhoDARTS: Differentiable Quantum Architecture Search with Density Matrix Simulations},
author={Swagat Kumar and Jan-Nico Zaech and Colin Michael Wilmott and Luc Van Gool},
year={2025},
eprint={2506.03697},
archivePrefix={arXiv},
primaryClass={quant-ph},
url={https://arxiv.org/abs/2506.03697},
}