School of Engineering Department of Electronic and Computer Engineering 167 Compact Models for Circuit Design Supervisor: SHAO Qiming / ECE Student: LU Yunyang / ELEC Course: UROP 1100, Fall The growing demands for data-intensive and energy-efficient computing tasks are quickly outpacing the computational capabilities of traditional von Neumann architectures. For instance, NP-hard problems are commonly encountered in areas such as combinatorial optimization, resource allocation, and finance. The computational time and hardware resources required to solve these problems increase exponentially with problem size, making them exceedingly difficult, or even impossible to solve within a practical time frame using conventional computers. Therefore, people come up with the new computer architecture based on the Magnetoresistive Random Access Memory (MRAM) for solving such problems. In this report, the aim is to explore how MRAM not only demonstrates ultra-low power consumption and high-speed data storage but also serves as compute-in-memory (CIM) cores for a range of optimization tasks. Specifically, this report will delve into its application to optimize the traveling salesman problem (TSP), one of the most prominent combinatorial optimization problems. EDA for Superconducting Quantum Computing Supervisor: SHAO Qiming / ECE Student: GAO Shiyuan / MATH-PMA Course: UROP 1100, Spring UROP 2100, Summer This paper explores the theory, implementation, and comparative analysis of optimization techniques for superconducting qubit control. We begin by deriving the foundational principles of the Krotov and GRAPE methods. For the CRAB method, we focus on its implementation with the Nelder-Mead algorithm, a gradient-free optimizer that iteratively refines a set of candidate controls. Our study includes a detailed examination of computational efficiency, convergence properties, and inherent challenges of each method. Key findings reveal distinct trade-offs: gradient-based methods ensure monotonic improvement but face local optima traps, while CRAB, though more exploratory, exhibits instability in convergence and computational cost scaling. Notably, we identify a unique “stable point” phenomenon in CRAB, where optimization stagnates at a predictable infidelity level, suggesting opportunities for hybrid strategies. By combining the rapid initial progress of CRAB with the precision of gradient-based refinements, we propose pathways to more efficient and reliable quantum control.
RkJQdWJsaXNoZXIy NDk5Njg=