Nature-inspired metaheuristics are very popular these days; their creation is usually justified based on the “no free lunch” theorem by Wolpert and MacReady. However, the creation of novel metaheuristics should be realized with care, not only for the sake of creation (cf. Sörensen reports on superficial metaheuristics); in other words, the inspiration should be solid and well-intended (it would be ideal if such methods were formally verified, however this happens very seldom, because of their complexity). In the case of the metaheuristics presented in this monograph, the inspiration comes from the works of Albert Bandura, a renowned Canadian-American psychologist. One of his most important contributions to contemporary science is the theory of social cognitive learning, showing that people do not only learn from their experiences (trial and error) but also by perceiving other people and (fortunately) their trials and errors. This saves a lot of effort, allowing us to utilize the knowledge gathered by perceiving others in order to build humankind’s self-knowledge. This inspiration leads to the proposal of a socio-cognitive metaheuristic paradigm consisting of the introduction or enhancement of the cognitive properties of particular metaheuristics. The most important achievements in this area are socio-cognitive Ant Colony Optimization and socio-cognitive Particle Swarm Optimization. The introduction of cognitive features into such computing algorithms allows us to reach better efficiency in solving selected hard benchmark problems. In this work, the above-mentioned novel algorithms are presented along with selected experimental results. Moreover, the socio-cognitive computing paradigm is defined, and the relationship of the selected metaheuristic algorithm to this paradigm is discussed. This metaphor is also considered as a reference for the selected classic and agent-based metaheuristics. These algorithms are identified by relating them to the literature background, and the possibilities of enhancing them with socio-cognitive features are discussed. Certain examples of further research are also identified. This monograph is meant to introducea novel perspective on the selected metaheuristics, defining the socio-cognitive computing paradigm and providing guidance in this area for readers who are interested in such nature-inspired computing methods.
- Spis treści
-
Abstract 7
Streszczenie 8
Preface 9
1. Social metaheuristics 13
1.1. Metaheuristics as methods of last resort 14
1.2. Social inspirations and agency in metaheuristics 16
1.3. Ant Colony Optimization 18
1.4. Particle Swarm Optimization 21
1.5. Evolutionary computing 24
1.6. Evolutionary agent-based computing 26
1.7. Estimation of Distribution Algorithms 31
1.8. Towards extension of social metaheuristics 35
2. From social to cognitive inspirations in computing systems 37
2.1. Perspective taking 37
2.2. Social Cognitive Theory 39
2.3. Social cognitive agent systems 41
2.4. Enhancing social metaheuristics with cognitive abilities 43
3. Socio-cognitive swarm metaheuristics 47
3.1. Socio-Cognitive Ant Colony Optimization 48
3.1.1. Selected hybrid metaheuristics based on ACO. 48
3.1.2. Multi-type ACO 49
3.1.3. Socio-Cognitive ACO 51
3.1.4. Selected experimental results 56
3.1.5. Emergence of population structure in Socio-cognitive ACO 60
3.1.6. Summary of Socio-Cognitive ACO research 67
3.2. Enhancing Particle Swarm Optimization with socio-cognitive inspirations 68
3.2.1. Selected hybrid metaheuristics based on PSO 69
3.2.2. From perspective taking to enhancing PSO 71
3.2.3. Socio-cognitively-inspired PSO 71
3.2.4. Experiments on Socio-Cognitive PSO 74
3.2.5. Adaptation of Population Structure in Socio-cognitive PSO 75
3.2.6. Summary of socio-cognitive PSO research 81
3.3. Socio-cognitive Stochastic Diffusion Search 82
3.4. Socio-cognitive swarm intelligence algorithms in light of Social Cognitive Theory 84
4. Socio-cognitive classic and EMAS-related hybrid metaheuristics 90
4.1. Parallel and co-evolutionary algorithms 91
4.2. Co-evolutionary EMAS metaheuristics 95
4.3. Clonal Selection Algorithm and immunological EMAS 98
4.4. Elitist EMAS for multi-objective optimization 102
4.5. Socio-cognitive COMMAop 104
4.6. Differential Evolution and hybrid EMAS/DE 107
4.7. EMAS and Particle Swarm Optimization 110
4.8. Cultural algorithm, memetic algorithm, and memetic EMAS 111
4.9. Classic metaheuristics in light of Social Cognitive Theory 117
4.10. Hybrid EMAS-related metaheuristics in light of Social Cognitive Theory 123
5. Summary 128
List of acronyms 132
Bibliography 133