Source-linked AI summary
A Brief Review of Nature-Inspired Algorithms for Optimization
Iztok Fister, Xin-She Yang, Iztok Fister, Janez Brest, Dušan Fister
TL;DR
Difficult real-world optimization problems motivate a review of nature-inspired metaheuristic algorithms and the limited understanding of their classifications and efficiency. The paper categorizes algorithms by inspiration source, surveys them across four broad groups, and concludes that some algorithms are more efficient and popular while further genuinely novel research is needed.
Problem
Real-world and NP-hard optimization problems are difficult, while algorithm classifications and the reasons for differing efficiency remain insufficiently understood.
Method
The paper reviews algorithms in the literature by grouping them into swarm intelligence, bio-inspired non-SI, physics/chemistry-based, and other categories.
Results
Some algorithms are more efficient and popular than others, and the review provides a broad information source for further research.
Takeaways & Limitations
Further studies should seek genuinely novel and useful algorithms that improve understanding of efficiency and address hard, large-scale real-world problems.
Takeaways & Limitations
The proposed classifications are not unique because algorithms may belong to multiple categories depending on the chosen focus or perspective.
Abstract
from arXiv · showhide
Swarm intelligence and bio-inspired algorithms form a hot topic in the developments of new algorithms inspired by nature. These nature-inspired metaheuristic algorithms can be based on swarm intelligence, biological systems, physical and chemical systems. Therefore, these algorithms can be called swarm-intelligence-based, bio-inspired, physics-based and chemistry-based, depending on the sources of inspiration. Though not all of them are efficient, a few algorithms have proved to be very efficient and thus have become popular tools for solving real-world problems. Some algorithms are insufficiently studied. The purpose of this review is to present a relatively comprehensive list of all the algorithms in the literature, so as to inspire further research.
1 INTRODUCTION
Real-world optimization problems, including NP-hard problems, are challenging and may require trial-and-error optimization techniques without guaranteeing an optimal solution. This review focuses on organizing newer algorithms by their sources of inspiration and presents an algorithm list across the paper’s sections.
- 1 INTRODUCTION: NP-hard optimization problems are challenging because efficient algorithms are unavailable and optimal solutions are not guaranteed.Optimization tools are therefore applied through trial-and-error approaches.
- 1 INTRODUCTION: Particle swarm optimization, cuckoo search, and firefly algorithm have gained popularity because of their high efficiency.
- 1 INTRODUCTION: Classifying about 40 algorithms systematically is difficult because classification criteria vary and lack an easy guideline.
- 1 INTRODUCTION: The review focuses on the source of inspiration rather than attempting a fully detailed classification by every possible characteristic.
- 1 INTRODUCTION: The paper analyzes inspiration sources, provides a list of algorithms, and concludes with suggestions for further work.
2 SOURCES OF INSPIRATION
Nature is treated as the broad source of inspiration for new algorithms, with biology providing the largest share and swarm intelligence forming a prominent subset. The review also includes physical, chemical, and other sources while acknowledging diverse algorithm origins.
- 2 SOURCES OF INSPIRATION: Nature is presented as a rich source of inspiration for developing new algorithms.
- 2 SOURCES OF INSPIRATION: Most nature-inspired algorithms draw on biological systems, making bio-inspired algorithms the largest fraction.
- 2 SOURCES OF INSPIRATION: Swarm-intelligence-based algorithms are a special class of bio-inspired algorithms based on collective behavior, including ant colony optimization and particle swarm optimization.
- 2 SOURCES OF INSPIRATION: The review also recognizes algorithms inspired by physical and chemical systems, and even by music.
- 2 SOURCES OF INSPIRATION: The paper groups inspiration sources into broad categories while noting that its categorization is not unique.
3 CLASSIFICATION OF ALGORITHMS
The review organizes nature-inspired algorithms into four broad categories, while noting that classifications can overlap and depend on the chosen perspective. It also describes a generic iterative search scheme and links efficiency to balancing exploration with local search.
- 3 CLASSIFICATION OF ALGORITHMS: The review divides algorithms into swarm intelligence, bio-inspired non-SI, physics/chemistry-based, and other categories.Well-established algorithms such as genetic algorithms are omitted because they are already widely known.
- 3 CLASSIFICATION OF ALGORITHMS: Classifications are not unique because algorithms may fit multiple categories depending on the emphasis, such as search trajectory, agent interaction, or updating equations.
- 3.1 Swarm intelligence based: Swarm intelligence uses interacting agents whose collective behavior can exhibit self-organization and collective intelligence.
- 3.1 Swarm intelligence based: SI-based algorithms use multiple agents inspired by social insects and animal societies, including the behaviors modeled by PSO and FA.
- 3.1 Swarm intelligence based: Information sharing, self-organization, co-evolution, learning, and easy parallelization help explain the popularity and practical use of SI-based algorithms.
- 3.2 Bio-inspired, but not SI based: SI-based algorithms form a subset of bio-inspired algorithms, which in turn form a subset of nature-inspired algorithms.
- 3.2 Bio-inspired, but not SI based: Some bio-inspired algorithms do not use swarming behavior, while DE can be difficult to classify because its biological connection is indirect.
- 3.3 Physics and Chemistry Based: Physics/chemistry-based algorithms mimic laws or systems such as electrical charges, gravity, and rivers, while other algorithms draw on social or emotional characteristics.
4 CONCLUSION
The review groups diverse nature-inspired algorithms into four categories and offers the resulting list as a starting point for further research, while cautioning that some algorithms are more efficient and popular than others. It also warns against creating minimally novel algorithms merely for publication, emphasizing genuinely useful studies for difficult problems.
- The review summarizes diverse algorithms in four categories as a basis for further research, while noting that the classification is not unique.The categories are swarm intelligence, bio-inspired but not swarm-intelligence-based, physics/chemistry-based, and others.
- Some algorithms are more efficient and popular than others, motivating additional study of their relative usefulness.
- The authors caution against inventing algorithms with little improvement and no novelty solely for publication.
- They aim to encourage genuinely novel and important research that helps solve hard, large-scale real-world problems.