Quantum-inspired adaptive mutation operator enabled PSO (QAMO-PSO) for parallel optimization and tailoring parameters of Kolmogorov–Arnold network

Particle swarm optimizer (PSO) is a biomimetic optimization algorithm well-known for its potential in addressing diversified optimization problems (OP). Although PSO is based on swarm-cognitive, it often suffers from attaining the global optima and the balance between exploration and exploitation, l...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:The Journal of supercomputing Ročník 81; číslo 14; s. 1310
Hlavní autoři: Agrawal, Umang Kumar, Panda, Nibedan
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York Springer US 06.09.2025
Springer Nature B.V
Témata:
Algorithms
Cognition
Compilers
Computer Science
Convergence
Datasets
Genetic algorithms
Interpreters
Methods
Mutation
Numerical analysis
Optimization algorithms
Particle swarm optimization
Physics
Processor Architectures
Programming Languages
Qubits (quantum computing)
Statistical analysis
Swarm intelligence
Quantum adaptive mutation operator
Kolmogorov–Arnold network
Qubit
Particle swarm optimization
Hybrid algorithm
ISSN:1573-0484, 0920-8542, 1573-0484
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Particle swarm optimizer (PSO) is a biomimetic optimization algorithm well-known for its potential in addressing diversified optimization problems (OP). Although PSO is based on swarm-cognitive, it often suffers from attaining the global optima and the balance between exploration and exploitation, leading to untimely convergence and compromised outcomes. To enhance such shortcomings, we have proposed the quantum adaptive mutation operator PSO (QAMO-PSO), where the QAMO is integrated with the standard PSO. In the suggested approach, the qubit dynamics and adaptable mutation concept are employed to amplify the search region of swarms, while preserving superior outcomes by altering the locations of swarms concerning the quanta rotation and including a dynamic quantum-inspired mutation mechanism. The QAMO frameworks concurrently modify the mutation likelihood regarding the global best fitness, adhering it to randomness to suppress the confined optima alongside accelerating optimal convergence. The competency of QAMO-PSO is rigorously quantified over the IEEE-CEC-2022 benchmark problem suite. The anticipated QAMO-PSO reveals commendable findings over the shifted and rotated (SR), hybrid and composition functions pertaining to CEC-2022. Furthermore, a cascade of statistical analysis is conducted to substantiate the distinctiveness of QAMO-PSO. Additionally, the applicability of QAMO-PSO is assessed in real-life constraints by tuning the hyperparameters of Kolmogorov–Arnold network (KAN), showcasing its efficacy in multivariate and nonlinear feature space. The QAMO-PSO as a trainer algorithm for KAN reveals prevailing results over the other baseline algorithms. Finally, the QAMO-PSO is a viable and reliable hybrid algorithm (HA) to tackle optimization problems related to engineering and computational cognition.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1573-0484
0920-8542
1573-0484
DOI:10.1007/s11227-025-07810-w