Dzahini presents scalable noise-aware optimizer for variational quantum algorithms
Kwassi Joseph Dzahini, an assistant computational mathematician in the Mathematics and Computer Science division at the U.S. Department of Energy’s Argonne National Laboratory, gave a talk at Sayas Numerics Day on May 9, 2026.
Hosted at the University of Maryland, College Park, Sayas Numerics Day brought together graduate students, postdocs and other early career researchers to share recent work in computational mathematics.
In his presentation “Noise-Aware Scalable Stochastic Trust-Region via Adaptive Random Subspaces” Dzahini introduced ANASTAARS, a stochastic derivative-free classical optimizer for the quantum approximate optimization algorithm (QAOA). The method addresses the increasing optimization difficulty that arises as the number of QAOA layers grows. Unlike earlier approaches involving the full parameter space, ANASTAARS adaptively selects smaller subspaces and reuses previous measurements, adding only a small number of new points when constructing the next interpolation model.
“With this strategy, we can reduce the number of new noisy function evaluations needed to build local models, which in the QAOA setting translates into fewer circuit-measurement batches,” Dzahini said.
To account for noisy function evaluations, ANASTAARS uses a trust-region framework, where a simplified local model is trusted only within a limited neighborhood of the current point and the size of that neighborhood is adjusted according to the model’s observed performance. The method further incorporates an estimate of the local noise level into the trust-region acceptance test, helping distinguish meaningful objective decrease from changes caused primarily by sampling noise.
Numerical experiments on QAOA benchmark problems show that the method maintains strong performance as the number of circuit layers, and hence the number of optimization parameters, increases.
“As QAOA circuits become deeper, the number of parameters to optimize grows, making the classical optimization step increasingly difficult. Our results suggest that ANASTAARS can help manage this challenge by using adaptive subspaces and noise-aware decisions.” Dzahini said.
For more details about ANASTAARS, see the full article “A Noise-Aware Scalable Subspace Classical Optimizer for the Quantum Approximate Optimization Algorithm” by Kwassi Joseph Dzahini, Jeffrey M. Larson, Matt Menickelly, and Stefan M. Wild, https://dx.doi.org/10.1109/QCE65121.2025.00013
Argonne National Laboratory seeks solutions to pressing national problems in science and technology by conducting leading-edge basic and applied research in virtually every scientific discipline. Argonne is managed by UChicago Argonne, LLC for the U.S. Department of Energy’s Office of Science.
The U.S. Department of Energy’s Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, visit https://energy.gov/science.