Expand QISE: Track 1: RLQSC: Reinforcement Learning for the Optimal Design of Programmable Quantum Sensor Circuit

Cleveland State University, Cleveland OH

Investigators

Abstract

Non-technical Description: The project aims to create and evaluate quantum and classical reinforcement learning-based agents for the optimal design of programmable quantum sensor circuit. The resulting technology from this research project will allow for better meters of the physical world with a breadth of applications that bridge many fields of science. The outcomes from this project will have a positive impact on the precision quantum-enhanced metrology measurements systems, such as the Noisy Intermediate Scale Quantum devices. The project will provide rich opportunities for Quantum Information Science and Engineering (QISE) research training and professional development. The project team will recruit, motivate, and train diverse and underrepresented minority and female students in QISE research methods. The project will have a positive impact on the K–16 QISE talent development pipeline and workforce development for the city of Cleveland, where underrepresented minorities such as African Americans and Hispanics constitute the majority of the population. Technical Description: Quantum sensing is a mature technology that has achieved remarkable progress over the past decades. The challenge going forward is to leverage potential gains from quantum entanglement and superposition to enable the next generation of sensors and thereby narrow the gap between the current performance and the fundamental limits set by quantum physics. However, the optimal design of a quantum sensor circuit that generates entangled qubits is a non-trivial task, which motivates the consideration of machine learning to assist with this design. Current efforts have in large part been limited to variational optimization of few parameter systems corresponding to simple circuits with few elements. To advance the state of the art, in this project, a reinforcement-learning-based optimal circuit design is developed for programmable quantum sensors. The specific objective is to create and evaluate quantum and classical reinforcement learning-based agents to design the deep circuit. The method utilizes a learning cycle of actions and rewards to generate the sequence of gates with optimal performance, using the measure of quantum Fisher information as a means to quantify the reward. The methodology involves multiple components such as demonstrations of the ideal system, evaluation of noise and imperfections, extensions to a quantum agent, and the performance evaluation of classical and quantum agents towards the design of the programmable quantum sensor circuit. Metrics used to evaluate the success of the research approach include sensitivity, dynamic range, robustness to dissipation and decoherence, and speed. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.