Source-linked AI summary
A swarm optimization algorithm inspired in the behavior of the social-spider
Erik Cuevas, Miguel Cienfuegos, Daniel Zaldivar, Marco Perez
TL;DR
The paper addresses limitations in swarm optimization, including premature convergence and imbalance between exploration and exploitation. It proposes Social Spider Optimization, modeling male and female spiders with gender-specific operators, and reports better results than PSO and ABC across benchmark functions.
Problem
Popular swarm algorithms such as PSO and ABC can prematurely converge, struggle with local minima, and damage the exploration-exploitation balance.
Method
SSO simulates cooperative social-spider colonies by dividing the population into male and female agents governed by different evolutionary operators.
Results
SSO delivered better results than PSO and ABC for all tested functions across a suite of 19 benchmark functions.
Takeaways & Limitations
Gender-specific operators and population division provide mechanisms associated with better particle distribution and different exploration-exploitation rates during evolution.
Abstract
from arXiv · showhide
Swarm intelligence is a research field that models the collective behavior in swarms of insects or animals. Several algorithms arising from such models have been proposed to solve a wide range of complex optimization problems. In this paper, a novel swarm algorithm called the Social Spider Optimization (SSO) is proposed for solving optimization tasks. The SSO algorithm is based on the simulation of cooperative behavior of social-spiders. In the proposed algorithm, individuals emulate a group of spiders which interact to each other based on the biological laws of the cooperative colony. The algorithm considers two different search agents (spiders): males and females. Depending on gender, each individual is conducted by a set of different evolutionary operators which mimic different cooperative behaviors that are typically found in the colony. In order to illustrate the proficiency and robustness of the proposed approach, it is compared to other well-known evolutionary methods. The comparison examines several standard benchmark functions that are commonly considered within the literature of evolutionary algorithms. The outcome shows a high performance of the proposed method for searching a global optimum with several benchmark functions.
Departamento de Electrónica
The paper is affiliated with Universidad de Guadalajara’s CUCEI and identifies swarm algorithms, global optimization, and bio-inspired algorithms as keywords.
- The listed affiliation is Universidad de Guadalajara, CUCEI, in Guadalajara, Jalisco, México.
- The paper’s keywords are swarm algorithms, global optimization, and bio-inspired algorithms.
1. Introduction
Swarm intelligence models collective animal behavior for distributed problem solving, while existing swarm optimizers face convergence and exploration challenges. The paper proposes SSO, which models gender-specific spiders and cooperative colony behavior for optimization.
- 1. Introduction: Swarm intelligence models collective behavior among insects or animals through individual interactions without centralized supervision.Its advantages include scalability, fault tolerance, adaptation, speed, modularity, autonomy, and parallelism.
- 1. Introduction: Self-organization and labor division let locally responsive units cooperate on global tasks while adapting to internal and external changes.
- 1. Introduction: Existing swarm optimizers include PSO, ABC, BFOA, and KH, which mimic bird, bee, bacterial, and krill behaviors.
- 1. Introduction: Most swarm algorithms treat individuals as unisex, missing selective operators based on characteristics such as sex or task responsibility.The paper links this limitation to opportunities for improving population diversity and search.
- 1. Introduction: PSO and ABC can prematurely converge or struggle with local minima because their position updates may distort the exploration-exploitation balance.
- 1. Introduction: Social-spider colonies use gender-dependent tasks and communal webs as communication channels for cooperative behavior.
- 1. Introduction: SSO models male and female spiders with different evolutionary operators that emulate cooperative colony behaviors and address flaws associated with PSO and ABC.
- 1. Introduction: The paper evaluates SSO experimentally and compares it with other evolutionary methods before presenting conclusions.
2. Biological fundamentals
Social-spider colonies coordinate through member interactions and a communal web, with behavior and responsibilities differentiated by gender. Their cooperation includes social interaction, reproduction, and task specialization.
- 2. Biological fundamentals: Social insect colonies are cooperative systems that self-organize under constraints and enable distributed operation and task specialization.
- 2. Biological fundamentals: Spiders are classified as solitary or social according to their level of cooperative behavior, with social spiders forming persistent colonies.
- 2. Biological fundamentals: A social-spider colony contains members and a communal web; members are divided into males and females who perform different activities.The web communicates information through vibrations used for collective coordination.
- 2. Biological fundamentals: The colony’s global cooperative patterns arise from internal interactions governed by social-interaction and mating rules.
- 2. Biological fundamentals: Female spiders show attraction or dislike toward colony members based on vibrations transmitted through the communal web.
- 2. Biological fundamentals: Male spiders are reproductive-oriented, with dominant and non-dominant classes; dominant males are attracted to nearby females.
- 2. Biological fundamentals: Mating supports colony survival and information exchange among members.
- 2. Biological fundamentals: When a dominant male finds females within a specific range, it mates with all of them to produce offspring.
3. The Social Spider Optimization (SSO) algorithm
SSO models optimization agents as gendered social spiders that communicate through a communal web and use specialized operators for exploration, exploitation, and mating. The algorithm assigns fitness-based weights, models vibration relationships, initializes gendered populations, and applies gender-specific cooperative behavior to address premature convergence and exploration–exploitation imbalance.
- Population structure: SSO divides the population into female and male spiders, with the female count randomly selected and the male count computed as its complement.The complete population is partitioned into groups F and M, preserving the total population size N.
- Fitness assignation: Each spider receives a weight representing the quality of its solution, calculated from its fitness relative to the population’s best and worst values.For maximization, worstS is the minimum fitness and bestS is the maximum fitness.
- Vibration modeling: The communal web transmits information through vibrations whose strength depends on the sender’s weight and distance from the receiving spider.SSO defines special relationships involving the nearest higher-weight spider, the globally best-weight spider, and the nearest female spider.
- Population initialization: The algorithm initializes female and male spider positions randomly and uniformly within prescribed lower and upper parameter bounds.The initialization equations use rand(0,1) independently across individuals and parameters.
- Gender-specific operators: Female spiders use attraction or repulsion movements based on perceived vibrations and a threshold, while specialized operators separately process agents by gender.This design targets premature convergence and incorrect exploration–exploitation balance identified in PSO and ABC.
4. Experimental results
SSO was evaluated on 19 benchmark functions against PSO and ABC under common experimental settings. It achieved better reported results across all functions, faster convergence in examples, and statistically significant differences.
- 19 benchmark functions were used to compare SSO with PSO and ABC.The experiments used populations of 50 individuals and a maximum of 1000 iterations.
- The comparison used PSO parameters c_1=2 and c_2=2, a linearly decreasing weight factor, ABC limit=100, and SSO PF=0.7.These settings were used for the reported algorithm comparisons.
- SSO delivered better results than PSO and ABC for all tested functions across the reported performance indexes.The indexes were Average Best-so-far, Median Best-so-far, and Standard Deviation of the best-so-far solution.
- SSO found the best solution in less than 400 iterations on average in the illustrated evolution curves, while the other algorithms required more iterations.The curves covered functions 1, 3, 5, 10, 15, and 19.
- All Wilcoxon test p-values were below 0.05, indicating statistically significant differences between SSO and the compared algorithms.The test used Average Best-so-far data at a 5% significance level.
5. Conclusions
The conclusion presents SSO as a gender-differentiated, cooperative social-spider optimization algorithm. Experiments on 19 benchmark functions reported acceptable solution quality, with performance associated with search-space distribution and differentiated exploration–exploitation rates.
- SSO models cooperative social-spider behavior using male and female search agents with different evolutionary operators.The operators emulate distinct cooperative behaviors within the colony.
- Unlike most swarm algorithms, SSO models individuals as two genders rather than unisex agents.The paper links this design to more realistic colony emulation and additional computational mechanisms.
- The design addresses premature convergence and incorrect exploration–exploitation balance identified as flaws in PSO and ABC.
- SSO was tested on 19 benchmark functions and compared with PSO and ABC, showing acceptable solution quality on all tested functions.
- The authors associate SSO’s performance with better particle distribution and different exploration–exploitation rates across population types.
Appendix A. List of benchmark functions
Appendix A lists the benchmark functions used in the experimental study, including their formulas, dimensionality, search ranges, and optimum information where provided.
- The listed benchmark definitions include function formulas, search intervals, and dimensions such as n=30.
- The appendix also contains functions expressed with bounded domains including [-100,100], [-10,10], [-500,500], and [-5.12,5.12].
- Named functions in the appendix include Sum of Squares, Zakharov, and Powelll.
- Appendix A presents the test functions used in the experimental study.
MATLAB SOFTWARE
The MATLAB software package provides SSO.m and five auxiliary functions for running the optimization implementation. The included example uses the Griewank function.
- The software contains SSO.m and five auxiliary functions: FeMove.m, Griewank.m, MaMove.m, Mating.m, and Survive.m.
- Users are instructed to place the files in a subdirectory and run SSO.m.
- SSO.m implements an optimization example using the Griewank function, which can be modified.