Introduction:
To tackle complex real world problems, scientists have been looking into natural processes and creatures - both as model and metaphor - for years. Optimization is at the heart of many natural processes including Darwinian evolution, social group behavior and foraging strategies. Over the last few decades, there has been remarkable growth in the field of nature-inspired search and optimization algorithms. Currently these techniques are applied to a variety of problems, ranging from scientific research to industry and commerce. The two main families of algorithms that primarily constitute this field today are the evolutionary computing methods and the swarm intelligence algorithms. Although both families of algorithms are generally dedicated towards solving search and optimization problems, they are certainly not equivalent, and each has its own distinguishing features. Reinforcing each other's performance makes powerful hybrid algorithms capable of solving many intractable search and optimization problems.

About the journal:
Swarm and Evolutionary Computation is the first peer-reviewed publication of its kind that aims at reporting the most recent research and developments in the area of nature-inspired intelligent computation based on the principles of swarm and evolutionary algorithms. It publishes advanced, innovative and interdisciplinary research involving the theoretical, experimental and practical aspects of the two paradigms and their hybridizations. Swarm and Evolutionary Computation is committed to timely publication of very high-quality, peer-reviewed, original articles that advance the state-of-the art of all aspects of evolutionary computation and swarm intelligence. Survey papers reviewing the state-of-the-art of timely topics will also be welcomed as well as novel and interesting applications.

Topics of Interest:
Topics of interest include but are not limited to: Genetic Algorithms, and Genetic Programming, Evolution Strategies, and Evolutionary Programming, Differential Evolution, Artificial Immune Systems, Particle Swarms, Ant Colony, Bacterial Foraging, Artificial Bees, Fireflies Algorithm, Harmony Search, Artificial Life, Digital Organisms, Estimation of Distribution Algorithms, Stochastic Diffusion Search, Quantum Computing, Nano Computing, Membrane Computing, Human-centric Computing, Hybridization of Algorithms, Memetic Computing, Autonomic Computing, Self-organizing systems, Combinatorial, Discrete, Binary, Constrained, Multi-objective, Multi-modal, Dynamic, and Large-scale Optimization.

Applications:
Furthermore, the journal fosters industrial uptake by publishing interesting and novel applications in fields and industries dealing with challenging search and optimization problems from domains such as (but not limited to): Aerospace, Systems and Control, Robotics, Power Systems, Communication Engineering, Operations Research and Decision Sciences, Financial Services and Engineering, (Management) Information Systems, Business Intelligence, internet computing, Sensors, Image Processing, Computational Chemistry, Manufacturing, Structural and Mechanical Designs, Bioinformatics, Computational Biology, Mathematical and Computational Psychology, Cognitive Neuroscience, Brain-computer Interfacing, Future Computing Devices, Nonlinear statistical and Applied Physics, and Environmental Modeling and Software.

Special Issue on Integration Method of Reinforcement Learning and Evolutionary Algorithm: Approaches and Applications
提出日: 2024-08-15

In recent years, a large number of problems have emerged in research and industry, that can be essentially classified as complex optimization problems. These problems are often characterized by complex features such as multi-modality, dynamics, discontinuity and nonlinearity. To date, numerous algorithms have been proposed to address such challenges, among which evolutionary algorithms (EAs) have garnered significant attention owing to their exceptional performance. However, the traditional EAs can result in high overhead due to excessive factor of randomness. To address this limitation, numerous methods have been proposed to enhance the algorithms from various perspectives.

Guest editors:
Dr. Yanjie Song (yanjiesong96@gmail.com)
Prof. Witold Pedrycz (wpedrycz@ualberta.ca)
Prof. Rammohan Mallipeddi (mallipeddi.ram@gmail.com)

Manuscript submission information:
Reinforcement learning (RL) has been identified as an effective approach for agents to learn the solution space by taking actions and interacting with the environment, thereby continuously updating their strategies. With the continuous evolution of methods, deep learning further enhances the potential of RL. Over the past three years, a substantial amount of research in EA improvement has focused on integrating RL into the EA framework, referred as reinforcement learning-assisted evolutionary algorithm (RL-EA). The RL-EA effectively leverages the acquired search information to collaboratively optimize solutions, demonstrating its success across various problem domains. Despite the successful application of RL-EA in many areas, the theoretical analysis of algorithms, benchmarks, training methods and strategy design is still an open field of research.Furthermore, a portion of the research aims to integrate EA into RL, known as evolutionary reinforcement learning (ERL). Within this algorithm framework, EA primarily handles tasks such as hyperparameter optimization, policy search, exploration, and reward shaping. The ERL is capable of handling large and complex RL tasks; however, it faces the challenge of high computational costs and sparse rewards. Therefore, there is a need to explore novel methods to enhance algorithm performance.Scope and topics:The main aim of this special issue is to report on the recent progress in integration methods of RL and EA (i.e. RL-EA, ERL). For RL-EA, special attention is paid to the theoretical analysis of algorithms, benchmarks and training methods specific to this new class of methods, and parameter tuning. This special issue also focuses on process design, algorithm framework, evaluation method, and benchmarks for ERL. In addition, it is anticipated that this special issue will deliver novel solution methods for some real-world problems, and propose some future direction, aiming to help readers gain a deeper comprehension of this field. We encourage the submission of original papers on topics of interest, including but not limited to the following:
RL-EA

Mechanisms for information interaction between RL and EA
Modelling of RL in RL-EA
Multi-agent reinforcement learning method design
Advance search strategies in RL-EA
Parameter control of EAs based on RL
Comparative studies with different types of RLs
Accelerated computing technologies for RL-EA
Training mechanism design of RL-EA
Time complexity analysis of RL-EA
Ensembles of different RLs in RL-EA
RL-EA for optimization
Hyperparameter tuning methods for RL-EA
Benchmark for RL-EA
Real-world applicationsERL1. Encoding of ERL2. Algorithm framework design3. EA for multi-agent RL4. EA for multi-objective RL5. EA for meta-RL6. Sampling and utilization7. Operators in ERL8. Evaluation methods9. Scalable platform design10. Benchmark for ERL11. ERL for complex sequential decision-making problems