CS Events Monthly View
Qualifying ExamEfficient Quantum Circuit Compilation with Permutable Operators through a Time-Optimal SWAP Insertion Approach |
|
||
Tuesday, May 09, 2023, 10:00am - 12:00pm |
|||
Abstract:
In the NISQ era of quantum computing, each qubit is physically connected to a limited number of neighboring qubits. This limitation requires a compiler to convert a quantum program into a hardware-compliant circuit, specifically, by mapping each two-qubit gate to two physical qubits connected by a link. Existing approaches use SWAP gates to dynamically remap logical qubits to physical qubits, but often lack consideration for time-optimality. We propose a theoretically time-optimal SWAP insertion scheme for the qubit mapping problem.
Later, the Quantum Approximation Optimization Algorithm (QAOA) gained attention as a highly advocated variational algorithm for solving combinatorial optimization problems. One critical feature of the QAOA quantum circuit is that it consists of permutable two-qubit operators that commute. The flexibility in ordering these permutable gates leads to shorter circuit depth but presents a challenge in circuit compilation due to the additional degree of freedom. To utilize this flexibility, we adapt our SWAP insertion scheme for generic circuits and propose a structured solution that considers connectivity regularity in scalable quantum hardware. We also adapt this structured method to practical cases in real hardware, addressing sparsity of input problem graphs and noise variability. We evaluate our method on IBM and Google architecture coupling graphs for up to 1,024 qubits and demonstrate better results in both depth and gate count, with up to a 72% reduction in depth and a 66% reduction in gate count. Our real experiments on IBM Mumbai show that we can find a better expected minimal energy than the state-of-the-art baseline.
Speaker: Yuwei Jin
Location : CoRE 305
Committee:
Professor Zheng Zhang (Advisor)
Professor Yipeng Huang
Professor Mario Szegedy
Professor Casimir Kulikowski
Event Type: Qualifying Exam
Abstract: See above
Organization:
Rutgers University
School of Arts & Sciences
Department of Computer Science
Contact Professor Zheng Zhang