The particle swarm optimization (PSO) algorithm is inspired by the social behaviors of biological populations, such as bird foraging, and performs optimization by simulating information exchange and collaboration among individuals within the swarm. However, how to fully utilize swarm information to effectively balance the exploration and convergence of the algorithm remains a key research issue. To this end, this paper proposes a Federated Learning Particle Swarm Optimization (FLPSO) algorithm to realize the parallelization and efficient collaboration of swarm information exchange. By primarily introducing the concept of federated evolutionary learning from federated learning, this algorithm designs a dynamic selection strategy to effectively screen the clients participating in the computation. Meanwhile, a dynamic aggregation strategy is adopted to extract valid local information and construct global elite particles to guide the update of the entire swarm. The parameters within the aforementioned mechanisms are all adaptively adjusted. Under the test sets of CEC 2013 and 2022, by comparing the variants of PSO in recent years and the Top algorithm in CEC, FLPSO shows better advantages. Meanwhile, through ablation and sensitivity experiments, the generalization and robustness of the proposed algorithm were further verified. For the relevant source code, please refer to: https://github.com/Donglin0730/FLPSO .
Membrane computing (P systems) provides biologically-inspired computational models capable of implementing complex algorithms through pattern matching with complex objects. Skeletonization, a fundamental step in pattern recognition, medical imaging, and fingerprint analysis, has been previously realized in P systems with complex objects using the Guo–Hall algorithm. This paper presents a publicly available Python simulator for the P system with complex objects implementing the Guo–Hall skeletonization algorithm. Experiments were run on Single-Process, Multi-Process, and Multi-Process with Numpy implementations of the simulator. It is shown in this work that the simulator is able to correctly simulate, based on empirical testings, the behavior of the P system implementing the Guo–Hall algorithm.
Membrane computing, particularly optimization spiking neural P systems (OSNPS), has emerged as a powerful paradigm for solving complex optimization problems. Despite its potential, existing OSNPS variants face significant challenges, including limited initial population diversity, premature convergence, and an insufficient balance between exploration and exploitation. These limitations hinder the algorithm’s ability to achieve optimal solutions efficiently and reliably, highlighting a critical gap in the current state of the art. To address these challenges, we propose An Optimization Spiking Neural P Systems with Synaptic Plasticity, named OSNPSsp, which integrates principles from biological systems and non-linear dynamics. Specifically, OSNPSsp incorporates two key innovations: a chaotic initialization strategy to enhance initial population diversity and a Hebbian learning-enhanced probability update mechanism to adaptively reinforce promising solutions. The theoretical significance of this work lies in the integration of non-linear dynamics and biological synaptic plasticity into the membrane computing framework, establishing a more principled approach to balancing exploration and exploitation. Practical evaluations on the 0/1 knapsack problem demonstrate that OSNPSsp significantly outperforms state-of-the-art algorithms, including GQA, NQEA, OSNPS, AOSNPS, and DAOSNPS. The results confirm that the proposed model not only delays premature convergence but also ensures superior solution quality and computational robustness, providing a scalable and efficient solution for complex combinatorial optimization challenges.
In this paper, we introduce a new indexation mechanism within a general class of P system models called kernel P systems, together with kPWorkbench, a software framework developed to support them. Our approach enables a compact definition of general patterns of rules, guards, objects, and compartments using syntactic constructs that are automatically expanded into fully specified kernel P systems. We also illustrate our approach through a case study, the Partition Problem.
Antimicrobial peptides (AMPs) offer promising alternatives to conventional antibiotics due to their broad-spectrum activity, low toxicity, and reduced risk of resistance development. However, the existing AMP design methods struggle to balance generation efficiency with pathogen-specific targeting, which limits their effectiveness against multi-pathogen infections. To address these challenges, we propose a novel AMP generation method termed Conditional Feedback Generative Adversarial Network (CFGAN), which integrates conditional information with a reinforcement feedback loop. Based on the GAN architecture, CFGAN incorporates two key innovations: (1) conditional information, encoding details about targeted pathogens, minimum inhibitory concentration level, and peptide length, which allows for precise control over antimicrobial specificity against ten classes of pathogens, and (2) reinforcement feedback loop, composing ten functional analysors against ten classes of pathogens and a dynamic reward mechanism that iteratively guides the generator to produce peptides with broad-spectrum targeting and high antimicrobial activity. To improve training stability and convergence speed, we implement a Brownian Motion Controller that dynamically adjusts the learning rates of both the generator and discriminator. Experimental results show that CFGAN outperforms existing AMP generation models in terms of consistency, stability, and antimicrobial properties. Additionally, molecular dynamics simulations confirm that the generated peptides possess strong structural stability and effective membrane-binding capacity. These findings demonstrate CFGAN as an efficient framework for designing broad-spectrum AMPs, with substantial implications for the development of next-generation anti-infective drugs. The data and Python code for CFGAN are available at https://github.com/YannanBin/CFGAN6 .
