Seminar: Application of AI-based methods to quantum metrology and quantum algorithm optimization

TITLE:  Application of AI-based methods to quantum metrology and quantum algorithm optimization   

AUTHOR: Sunny Xin Wang Department of Physics, City University of Hong Kong, Hong Kong SAR, China 

INVITED BY: Xi Chen

WHEN: July, 28 - 12PM

WHERE: Salón de Actos, ICMM

Abstract:

Recent advances in Artificial Intelligence have profoundly influenced science and technology, in particular quantum technologies. In this talk, we present our recent works illustrating how reinforcement learning (RL) and physics-informed neural networks (PINN) can be applied to quantum metrology and quantum algorithm optimization while improving resource efficiency across the quantum workflow.  First, we proposed RL algorithm which can discover generalizable quantum control strategies for parameter estimation tasks. By learning policies that remain effective across a range of parameter values, it enables adaptive control protocols that avoid repeated recalculation of optimal strategies during iterative estimation procedures [1]. Building on this idea, we introduce a reinforcement-learning-based framework for automated circuit architecture discovery in variational quantum sensing. By treating circuit structure as a trainable object and enabling distributed architecture search under hardware constraints, the method adaptively explores state preparation and measurement designs, achieving improved sensitivity and resource efficiency compared with fixed ansatzes [2]. We then present PINN approaches that enhance quantum system characterization and variational algorithm optimization. For Hamiltonian learning, we propose inverse physics-informed neural networks that incorporate the Schrödinger equation into the training objective, enabling accurate inference of system parameters from limited noisy measurements while maintaining physical consistency [3]. Finally, for noisy variational quantum algorithms, we introduce a physics-informed denoising network that learns the effective optimization dynamics under an error-mitigated cost landscape, significantly reducing the number of required circuit executions while preserving the correct optimization trajectory [4]. These results illustrate how AI-driven approaches can improve efficiency across multiple layers of the quantum computing stack and highlight the growing role of AI as a key enabler for scalable quantum technologies.  

References:

[1] Generalizable control for quantum parameter estimation through reinforcement learning, H. Xu, J. Li, L, Liu et al., npj Quantum Inf., 5, 82, 2019. 

[2] Automating Variational Quantum Sensing through Reinforcement-Learned Circuit Structures, J. Liu, X. Wang, in preparation. 

[3] Hamiltonian learning via inverse physics-informed neural networks, J. Liu, X. Wang, Phys. Rev. Research, 7, 043137, 2025. 

[4] Accelerating Noisy Variational Quantum Algorithms with Physics-Informed Denoising Networks, J. Liu, X. Wang, in preparation.