Optimization Strategies for Variational Quantum Algorithms in Noisy Landscapes

Vojtěch Novák, Ivan Zelinka, Václav Snášel · 2025

Read Optimization Strategies for Variational Quantum Algorithms in Noisy Landscapes by Vojtěch Novák, Ivan Zelinka, Václav Snášel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-01715.txt.

Introduction

Vojtech Novak, Ivan Zelinka, and Vaclav Snasel bench in 2025 more than fifty search heuristics for a quantum solver on near term gear. Such solvers face loud data, flat dead zones, and bent fields. The test solver seeks the low energy state of a quantum system.

A three phase sieve runs from a plain spin test to a nine bit scale test, then to a big model with near two hundred knobs. Noise turns smooth bowls rough, which kills plain slope tools. Two evolved swarm forms won in a steady way across all phases. Star tools of the past fell off hard with noise.

Catalog record

CREATORVojtěch Novák, Ivan Zelinka, Václav Snášel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE113 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-01715

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Optimization Strategies for Variational Quantum Algorithms in Noisy Landscapes Vojtěch Novák, Ivan Zelinka, Václav Snášel Abstract Variational Quantum Algorithms (VQAs) are a promising tool in the NISQ era, leveraging quantum computing across diverse fields. However, their performance is hindered by optimization challenges like local minima, barren plateaus, and noise from current quantum hardware. Variational Quantum Eigensolver (VQE), a key subset of VQAs, approximates molecular ground-state energies by minimizing a Hamiltonian, enabling quantum chemistry applications. Beyond this, VQE contributes to condensed matter physics by exploring quantum phase transitions and exotic states, and to quantum machine learning by optimizing parameterized circuits for classifiers and generative models. This study systematically evaluates over 50 meta-heuristic optimization algorithms—including evolution-based, swarm-based, and music-inspired methods—on their ability to navigate VQE’s multimodal and noisy landscapes. Using a multi-phase sieve-like approach, we identify the most capable optimizers and compare their performance on a 1D Ising model (3–9 qubits). Further testing on the Hubbard model (up to 192 parameters) reveals insights into convergence rates, effectiveness, and resilience under noise, offering valuable guidance for advancing optimization in noisy quantum environments. keywords Variational quantum algorithms, VQE, Optimization, Metaheuristics 1 Introduction Applications such as simulating intricate quantum systems or tackling vast linear algebra problems pose formidable challenges for classical computers, primarily due to their exorbitant computational demands. While fault-tolerant quantum computers may not materialize in the immediate future, quantum computing holds promise as a transformative solution. However, current quantum devices are constrained by significant limitations, including a restricted number of qubits and noise-induced constraints on circuit depth. Variational Quantum Algorithms (VQAs) have emerged as a leading approach to navigate these challenges. By employing a classical optimizer to refine a parametrized quantum circuit, VQAs offer a pathway to harness the power of quantum computation within existing constraints, despite the inherent noise in these devices. A key application of VQE, a subset of VQAs, is approximating the ground-state energy of molecular systems in quantum chemistry. While not a direct replacement for classical methods like Hartree-Fock, Configuration Interaction (CI), or Coupled Cluster (CC) theory, VQE provides a quantum-native alternative. However, current implementations often depend on classical preprocessing steps and remain experimental for most practical problems. The computational advantage of VQE is expected to become evident as quantum hardware advances, enabling efficient handling of larger, more complex molecular systems. Beyond quantum chemistry, VQAs have demonstrated potential in condensed matter physics, facilitating the study of quantum phase transitions and exotic states of matter. In quantum machine learning, parametrized quantum circuits optimized by VQAs enhance model performance in classification and generative tasks. Additionally, VQAs are applied in solving combinatorial optimization problems, quantum metrology, and quantum simulations relevant to material science. These diverse applications underscore the adaptability of VQAs to leverage quantum computing across multiple disciplines. This work focuses on a comprehensive comparative analysis of meta-heuristic algorithms for optimizing Variational Quantum Eigensolver (VQE) problems. Our study spans bio-inspired, physics-based, music-based, human-based, swarm-based, system-based, math-based, and evolution-based algorithms. Traditional gradient-based methods, such as Simultaneous Perturbation Stochastic Approximation (SPSA) and Constrained Optimization BY Linear Approximations (COBYLA), are also evaluated. However, these methods often struggle to locate global minima in the VQE landscape, especially under noisy conditions. Swarm-based algorithms, such as Particle Swarm Optimization (PSO) and Self-Organizing Migrating Algorithm (SOMA), mimic collective behavior in nature. Evolutionary-inspired algorithms, including Differential Evolution (DE) and Genetic Algorithms (GA), rely on principles of natural selection and adaptation. Other approaches, such as Simulated Annealing (SA) and Covariance Matrix Adaptation Evolution Strategy (CMA-ES), leverage physical or statistical principles for optimization. By conducting a thorough evaluation, this research aims to identify the most effective optimization strategies for VQE in noisy intermediate-scale quantum (NISQ) devices. Addressing challenges like barren plateaus, noise-induced local minima, and stochastic quantum measurements, we explore each algorithm’s robustness, scalability, and adaptability. This work seeks to advance the practical deployment of VQAs in quantum chemistry and beyond, bridging the gap between current limitations and the broader potential of quantum computing. 2 Variational quantum algorithms Quantum Machine Learning (QML) in the noisy intermediate-scale quantum (NISQ) era represents a paradigm shift, driven by the recognition that current quantum hardware is characterized by a limited number of noisy qubits (typically ranging from 20 to 1000). These systems face significant challenges in implementing deep quantum circuits due to the presence of noise [ 1 ] . The introduction of parameterized quantum circuits in 2014 catalyzed the development of QML, with notable algorithms such as the Quantum Approximate Optimization Algorithm (QAOA) [ 2 ] and the Variational Quantum Eigensolver (VQE) [ 3 ] laying the groundwork for this field. Both algorithms fall under the category of variational quantum algorithms (VQAs), which employ parameterized quantum circuits (PQCs) optimized through classical methods. These hybrid protocols combine the strengths of quantum and classical computing, with the quantum processor functioning as a trainable model refined through iterative sample-based measurements. VQAs are central to the NISQ era, offering a framework for solving a variety of problems through hybrid quantum-classical optimization loops [ 4 ] . In the context of QML, the term ”quantum neural networks” is often used to describe the application of PQCs in machine learning tasks [ 5 , 6 ] . These models aim to harness the representational power of quantum systems for tasks such as supervised learning and generative modeling [ 7 ] . Ongoing research focuses on enhancing the performance and scalability of VQAs [ 8 , 9 , 10 ] . However, whether QML algorithms can surpass classical counterparts remains an open question [ 11 , 12 , 13 ] . Significant strides have been made in circuit design, with diverse architectures proposed to optimize PQCs. These include hardware-efficient circuits [ 14 ] , the quantum alternating operator ansatz [ 15 ] , and dissipative quantum neural networks (dQNNs) [ 16 ] . Techniques like data re-uploading have further demonstrated the potential to enhance VQA performance, particularly for encoding classical input data into quantum models [ 17 ] . The hybrid quantum-classical paradigm has been effectively applied to simplified problems across various domains 1 , including quantum chemistry, combinatorial optimization, and machine learning. For example, the VQE has been used to approximate the ground state of molecular electronic Hamiltonians, offering a quantum-native alternative to classical methods for certain molecular systems. Similarly, the QAOA addresses combinatorial problems, such as finding approximate solutions to MaxCut and classical Ising models. These applications illustrate the versatility of hybrid algorithms in leveraging the strengths of current quantum hardware. Despite their promise, VQAs face challenges such as noise-induced barren plateaus and the stochastic nature of quantum measurements, which can complicate optimization. Nonetheless, as quantum hardware and software continue to evolve, experimental demonstrations are increasingly showcasing the practical potential of VQAs [ 7 ] . This rapidly advancing field holds the promise of enabling real-world applications in domains such as material science, drug discovery, and artificial intelligence. Figure 1: Applications of Variational Quantum Algorithms. source: https://arxiv.org/pdf/2012.09265.pdf Ising Model Without Magnetic Field The 1D Ising model, in the absence of an external magnetic field, is a cornerstone of statistical mechanics and quantum computing. Its simplicity and nontrivial solution space provide valuable insights into collective phenomena and optimization landscapes. This model is particularly favored as a benchmark for testing quantum algorithms,

Read the complete text