Noise-Aware ANASTAARS Optimizer Scales Quantum Approximate Optimization Algorithms Efficiently

The quest to harness the power of quantum computers for practical problem-solving faces a significant hurdle: optimising the complex calculations required to run quantum algorithms. Researchers at Argonne National Laboratory, led by K. J. Dzahini, J. M. Larson, and M. Menickelly, alongside S. M. Wild from Lawrence Berkeley National Laboratory, present a new method to address this challenge within the Quantum Approximate Optimisation Algorithm (QAOA). Their approach, called ANASTAARS, dramatically reduces the computational cost of optimising QAOA circuits, particularly as the complexity of these circuits increases, by intelligently selecting and reusing previous calculation results. Crucially, ANASTAARS also incorporates techniques to mitigate the impact of noise, a pervasive issue in current quantum hardware, paving the way for more reliable and scalable quantum computations on near-term devices.

QAOA Parameter Optimization with Classical Tools

Scaling Up Quantum Optimization with Smarter Classical Tools The development of practical quantum computers faces a significant challenge: current devices are limited in scale and prone to errors. This has driven interest in variational quantum algorithms (VQAs), which combine the strengths of quantum and classical computing to solve complex problems. VQAs utilize a quantum computer to prepare a solution and a classical computer to refine it, iteratively improving the result. The Quantum Approximate Optimization Algorithm (QAOA) is a prominent example, holding promise for tackling difficult optimization problems in fields like logistics, finance, and materials science.

However, optimizing the parameters within QAOA is a substantial hurdle. The process involves navigating a complex landscape of possible solutions, complicated by the inherent noise in quantum hardware. Traditional optimization methods, relying on gradients, can be computationally expensive and sensitive to this noise, potentially getting stuck in suboptimal solutions. Derivative-free optimization methods offer an alternative, but their performance often diminishes as the complexity of the problem and the number of quantum circuit layers increase, limiting the creation of deeper, more powerful circuits.

To address this, researchers developed ANASTAARS, a novel classical optimizer specifically designed for QAOA. ANASTAARS employs an adaptive strategy that intelligently explores the solution space by focusing on promising regions and reusing previously acquired data. By constructing simplified models within low-dimensional spaces, the algorithm efficiently identifies optimal parameters while minimizing the number of costly quantum measurements. Crucially, ANASTAARS also incorporates techniques to estimate and account for noise, making it more robust in the presence of imperfect quantum hardware. The innovation of ANASTAARS lies in its ability to scale effectively with increasing problem complexity.

By selectively reusing data and adapting to noise, the algorithm overcomes the limitations of existing methods, enabling the optimization of QAOA circuits with a greater number of layers. This advancement is critical for unlocking the full potential of near-term quantum computers and paving the way for practical applications of quantum optimization in a variety of fields. The team’s work represents a significant step towards harnessing the power of quantum computing, not by solely focusing on quantum hardware, but by developing smarter classical tools to support it.

Optimizing QAOA with Randomized Subspace Algorithms

This research details a new optimization algorithm for Variational Quantum Algorithms (VQAs), comparing its performance to several other methods. The core problem is that VQAs rely on classical optimization algorithms to adjust the parameters of the quantum circuit, and this classical optimization often becomes a bottleneck. Calculating gradients for VQAs can be expensive or noisy, so derivative-free optimization methods are often preferred, but these methods struggle to scale to larger, more complex quantum circuits. The proposed algorithm, ANASTAARS-QD2, utilizes an adaptive subspace strategy, intelligently exploring different subspaces of the parameter space and focusing on promising regions.

It is a randomized algorithm, helping to avoid getting stuck in local optima, and can increase the dimensionality of the explored subspace if progress stalls. Experiments focused on the Quantum Approximate Optimization Algorithm (QAOA) and tested the algorithm on both simple and challenging benchmark problems. Performance was measured by evaluating the optimization trajectory, the final solution quality, scalability, and the number of quantum circuit executions required. ANASTAARS-QD2 achieved performance comparable to a state-of-the-art optimization algorithm, avoided getting stuck in local optima, and demonstrated good scalability as the number of parameters increased. The results suggest that ANASTAARS-QD2 could enable the exploration of larger and more complex QAOA problems than are currently feasible, representing a significant step forward in the development of efficient optimization algorithms for VQAs and potentially paving the way for more powerful quantum algorithms with practical implications for fields like finance, machine learning, and materials science.

ANASTAARS-QD2 Optimizes Variational Quantum Algorithms

Random embeddings are being explored as a method to improve the performance of quantum optimization algorithms. These techniques aim to enhance the ability of algorithms to navigate complex solution spaces and find optimal solutions. Research indicates that these embeddings can contribute to more efficient and effective quantum optimization processes.