Adam assisted Fully informed Particle Swarm Optimzation ( Adam-FIPSO ) based Parameter Prediction for the Quantum Approximate Optimization Algorithm (QAOA)
SHASHANK SANJAY BHAT 1, PEIYONG WANG 1,2 , AND UDAYA PARAMPALLI. 1 · 2025
Read Adam assisted Fully informed Particle Swarm Optimzation ( Adam-FIPSO ) based Parameter Prediction for the Quantum Approximate Optimization Algorithm (QAOA) by SHASHANK SANJAY BHAT 1, PEIYONG WANG 1,2 , AND UDAYA PARAMPALLI. 1 (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-2506-06790.txt.
Introduction
Shashank Sanjay Bhat, Peiyong Wang, and Udaya Parampalli benchmark swarm methods for warming up a quantum optimizer, in this 2025 paper. The quantum approximate optimization algorithm solves hard either-or problems, but only when its dials start near good settings. Finding those settings is itself an optimization problem. The team tests particle swarm variants, where candidate solutions drift through the parameter space sharing what they learn.
One variant borrows the Adam optimizer's adaptive steps to guide the swarm. Weighted MaxCut problems of several sizes, circuit depths, and noise levels form the test bed. Good initialization decides whether near-term quantum hardware spends its precious runs solving or wandering. Benchmarks like this tell practitioners which classical helper earns its keep.
Catalog record
| CREATOR | SHASHANK SANJAY BHAT 1, PEIYONG WANG 1,2 , AND UDAYA PARAMPALLI. 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-06790
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Adam assisted Fully informed Particle Swarm Optimzation ( Adam-FIPSO ) based Parameter Prediction for the Quantum Approximate Optimization Algorithm (QAOA) SHASHANK SANJAY BHAT 1, PEIYONG WANG 1,2 , AND UDAYA PARAMPALLI. 1 Abstract The Quantum Approximate Optimization Algorithm (QAOA) is a prominent variational algorithm used for solving combinatorial optimization problems such as the Max-Cut problem. A key challenge in QAOA lies in efficiently identifying suitable parameters ( γ , β ) (\gamma,\beta) that lead to high-quality solutions. In this paper, we propose a framework that combines Fully Informed Particle Swarm Optimization (FIPSO) with adaptive gradient correction using the Adam Optimizer to navigate the QAOA parameter space. This approach aims to avoid issues such as barren plateaus and convergence to local minima. The proposed algorithm is evaluated against two classes of graph instances, Erdos Renyi and Watts-Strogatz. Experimental results across multiple QAOA depths consistently demonstrate superior performance compared to random initialization, underscoring the effectiveness and robustness of the proposed optimization framework. † † doi: 10.1109/TQE.2020.DOI † † address: School of Computing and Information Systems, the University of Melbourne, Parkville VIC 3010, Australia † † address: CSIRO Clayton, Research Way Clayton VIC 3168, Australia † † corresponding: Corresponding author: Shashank Sanjay Bhat (email: [email protected]). I Introduction In recent years, quantum computing has emerged as a promising approach to solving complex optimization problems such as TSP (Traveling Salesman Problem) [ 1 ] , VRP (Vehicle Routing Problem) [ 2 ] , and the Max-Cut problem [ 3 ] . We are currently in the NISQ (Noisy Intermediate Scale Quantum) era [ 4 ] , characterized by devices with limited numbers of qubits that are susceptible to noise and decoherence. These constraints have inspired the development of hybrid quantum-classical algorithms, which leverage the strengths of both paradigms. In such frameworks, quantum circuits handle computationally intensive tasks, while classical optimizers iteratively refine results [ 5 ] . One such hybrid quantum-classical algorithm is QAOA [ 6 ] , which is particularly suited for tackling combinatorial optimization problems like the Max-Cut problem. Max-Cut problem is a NP-hard problem that involves partitioning a graph’s vertices into two subsets to maximize the number of edges between them. QAOA operates by preparing candidate solutions on a quantum circuit using parameters \gamma and \beta . A classical optimizer then adjusts these parameters to improve the solution quality [ 8 ] . Finding optimal parameters for QAOA can be challenging due to issues like barren plateaus [ 7 ] and local minima in the optimization landscape [ 10 ] . To address these challenges, effective warm-start methods are crucial for ensuring that the algorithm begins its search near promising regions of the solution space. To address these challenges in parameter optimization for QAOA, we propose using Adam-assisted Fully Informed Particle Swarm Optimization (Adam-FIPSO) to predict near-optimal parameters. By combining a custom objective function with Adam-FIPSO, our method effectively searches for parameters that balance performance and optimization efficiency. The key components of our approach are listed below: • Custom Objective Function: We introduce a custom objective function that aims to achieve near-optimal performance while incorporating the Approximation Ratio (AR). This function helps in guiding the optimization process to improve the quality of the predicted solution. • Adam-Assisted FIPSO for Parameter Prediction: We employ the Adam-FIPSO algorithm to predict near-optimal QAOA parameters \gamma_{1},\gamma_{2},\gamma_{3} and \beta_{1},\beta_{2},\beta_{3} for varying circuit depths ( p=1,2,3 ). This optimization approach enhances the search process by utilizing the collective knowledge of the swarm. • Improved Approximate Ratios: Evaluations using the approximate ratio parameter (comparing the quality of the cut found with the best possible cut) demonstrate that our Adam-FIPSO-based initialization significantly outperforms random initialization. II Related Work Several recent studies have investigated machine learning techniques to improve parameter initialization in the Quantum Approximate Optimization Algorithm (QAOA) for combinatorial problems such as Max-Cut. Liang et al. [ 13 ] employ Graph Neural Networks (GNNs) to learn mappings from graph structures to effective QAOA parameters, aiming to reduce reliance on iterative quantum evaluations. Similarly, Meng et al. [ 14 ] introduce a denoising diffusion probabilistic model (DDPM) to generate high-quality parameter samples that serve as strong initializations. While these methods are model-driven and rely on training over large graph datasets, they do not explicitly utilize the QAOA cost landscape during optimization. In contrast, our approach directly optimizes QAOA parameters by evaluating the expectation value of the QAOA circuit for each graph instance. This is achieved using an Adam-assisted Fully Informed Particle Swarm Optimization (Adam-FIPSO) strategy, which combines swarm-based exploration with gradient-based refinement. By operating directly on the QAOA objective function without the need for learned surrogates, our method provides a flexible and model-free alternative for parameter tuning across varying graph types and circuit depths. III Preliminaries III-A QAOA QAOA is a variational quantum algorithm designed to solve combinatorial optimization problems by leveraging quantum mechanics [ 6 ] . It operates by encoding an optimization problem into a quantum circuit consisting of alternating cost and mixer Hamiltonians, whose parameters are optimized to approximate the optimal solution. Fig. 1: The QAOA Circuit mechanism showing the cost and mixer circuit components QAOA applies a sequence of unitary transformations using the cost and mixer Hamiltonians. Given a depth parameter p , the quantum state evolves as: \displaystyle\ket{\psi_{p}(\gamma,\beta)}=e^{-i\beta_{p}H_{M}}e^{-i\gamma_{p}H_{C}} (1) \displaystyle\dots e^{-i\beta_{1}H_{M}}e^{-i\gamma_{1}H_{C}}\ket{+}^{\otimes n}. • The cost Hamiltonian evolution e^{-i\gamma H_{C}} encodes problem constraints. • The mixer Hamiltonian evolution e^{-i\beta H_{M}} introduces transitions between states. • The initial state \ket{+}^{\otimes n} is a uniform superposition: \ket{+}=\frac{1}{\sqrt{2}}(\ket{0}+\ket{1}). (2) The circuit alternates between these operators for p layers, with variational parameters \gamma=(\gamma_{1},\dots,\gamma_{p}) and \beta=(\beta_{1},\dots,\beta_{p}) optimized to minimize the expectation value of H_{C} : \langle H_{C}\rangle=\bra{\psi_{p}(\gamma,\beta)}H_{C}\ket{\psi_{p}(\gamma,\beta)}. (3) III-B Max Cut Problem The Max-Cut problem is a well-known combinatorial optimization problem where the goal is to partition the vertices of a given graph G(V,E) into two sets such that the number of edges crossing the partition is maximized. Formally, for a graph G with |V|=n vertices and edges (i,j)\in E , the problem can be expressed as maximising the objective function: C(z)=\sum_{(i,j)\in E}\frac{1}{2}(1-z_{i}z_{j}), (4) where z_{i}\in\{-1,1\} represents the assignment of vertex i to one of the two partitions. The corresponding cost Hamiltonian in QAOA is: H_{C}=\sum_{(i,j)\in E}\frac{1}{2}(1-Z_{i}Z_{j}), (5) where Z_{i} is the Pauli-Z operator acting on the i th qubit. The ground state of H_{C} corresponds to the optimal Max-Cut partition. IV Methodology IV-A Problem Definition This work addresses the problem of optimizing the variational parameters \boldsymbol{\gamma}=(\gamma_{1},\gamma_{2},\gamma_{3}) and \boldsymbol{\beta}=(\beta_{1},\beta_{2},\beta_{3}) in the Quantum Approximate Optimization Algorithm (QAOA) for solving the MaxCut problem over diverse graph instances. Given a set of graphs (G_{1},G_{2},\dots,G_{n}) , the goal is to identify high-quality QAOA parameters that yield near-optimal cut values for each instance and depth. However, the QAOA optimization landscape is known to be highly non-convex, oscillatory, and problem-dependent, making traditional gradient-free methods prone to premature convergence or stagnation in local optima. Conventional Particle Swarm Optimization (PSO) techniques, while effective in simpler domains, often struggle to efficiently explore such complex energy landscapes due to limited adaptation to local curvature. Fully Informed PSO (FIPSO) partially mitigates this by incorporating social knowledge from all particles, but still lacks fine-grained control over per-dimension learning dynamics. In contrast, the proposed Adam-assisted Fully Informed PSO (Adam-FIPSO) augments the social guidance of FIPSO