Seminar Announcement: Monday, 26 October 2020, 11:00 AM EDT (online)
A dissertation on
Benchmarks and Controls for Optimization with Quantum Annealing
Presented by
Ms. Erica Grant
Bredesen Center for Interdisciplinary Research and Graduate Education
University of Tennessee, Knoxville
Monday, 26 October 2020, 11:00 AM EDT (online)
Abstract: Quantum annealing (QA) is a metaheuristic specialized for solving optimization problems which uses principles of Adiabatic Quantum Computing (AQC), namely the adiabatic theorem. There are devices which implement QA using quantum mechanical phenomena. These QA devices don’t perfectly adhere to the adiabatic theorem because they are subject to thermal and magnetic noise. Thus, QA devices return statistical solutions with some probability of success where this probability is impacted by the level of noise of the system. As these devices improve, it is believed that they will become less noisy and more accurate. However, there are tuning strategies that may further improve that probability of finding the correct solution and reduce the effects of noise on solution outcome. In this dissertation, these tuning strategies are explored in depth to determine the impact of pre-processing, annealing, and post-processing controls on performance. In particular, these tuning strategies are applied to a real-world NP-Hard optimization problem, portfolio optimization. Although there was very little performance improvement from tuning the spin reversal transforms, anneal time, and embedding, the results revealed that reverse annealing controls improved the probability of success by an order of magnitude over forward annealing alone. The chain strength experiments revealed that increasing the strength of the intra-chain coupling improves the probability of success until the intra-chain coupling strengths begin to overpower the inter-chain couplings. By taking a closer look at which physical qubits in the embedded chains, the probability for each qubit to be faulty was visualized and was used to develop a post-processing strategy that outperformed the standard which chooses a logical qubit value from a broken chain. The results of these findings provide a guide for current researchers to find the optimal set of controls for their unique real-world optimization problem to determine whether QA provides some benefit over classical computing, lay the ground-work for developing new tuning strategies that could further improve performance, and characterize the current hardware for benchmarking future generations of QA hardware.
Speaker Biography: Erica Grant grew up in Richmond, VA and has a BS in Physics and minor in Nanoscience from Virginia Tech. She became interested in quantum computing after two internships at Oak Ridge National Laboratory. Since 2016, she has studied quantum computing at the University of Tennessee’s Bredesen Center where her research focuses on tuning strategies and benchmarks for quantum annealing.
Please contact Travis Humble, humblets@ornl.gov, for any questions regarding the seminar presentation.
Online via Zoom
Leave a Reply