580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can evolutionary algorithm parameters be adaptively controlled to balance exploration and exploitation for improved optimization performance?
This research area focuses on the critical challenge of parameter configuration within evolutionary algorithms (EAs) such as genetic algorithms (GAs), evolutionary strategies, and related metaheuristics. Properly balancing exploration (diversification) and exploitation (intensification) through dynamic parameter control—like mutation rate, crossover probability, population size, and selection pressure—is vital to avoid premature convergence and enhance solution quality. Adaptive parameter control methods aim to optimize these values automatically during the evolutionary process, leveraging feedback mechanisms to adjust parameters in response to algorithm performance, thereby improving convergence rates and robustness across diverse problem instances.
Key finding: Proposes a novel adaptive parameter control method that repeatedly redefines algorithm parameters via a separate optimization process using recent performance feedback. This method predicts promising parameter values for... Read more
Key finding: Introduces the Genetic Algorithm Based on Natural Selection Theory (GABONST), an enhanced GA variant designed to improve control over exploration and exploitation phases by modifying the natural selection process and mutation... Read more
Key finding: Develops a hybrid genetic algorithm (HGA) integrating GA for global search with particle swarm optimization (PSO) for local search to effectively address multi-objective optimization problems. The hybrid approach adaptively... Read more
2. What are the advantages and design principles of novel bio-inspired evolutionary algorithms beyond classical genetic algorithms for complex optimization tasks?
This theme explores the development and application of new bio-inspired evolutionary optimization algorithms drawing from diverse natural processes beyond classical genetic paradigms. These include mechanisms from biological organ stability (allostasis), animal foraging, and social behaviors of species such as egrets and red pandas. The focus is on leveraging novel metaphors and evolutionary operators to improve exploration-exploitation trade-offs, maintain population diversity, avoid local optima, and adapt dynamically to problem landscapes. These bio-inspired strategies contribute fresh algorithmic components and parameter-free models that enhance performance on benchmark problems and real-world applications.
Key finding: Proposes Allostatic Optimisation (AO), a novel evolutionary algorithm inspired by biological allostasis, modeling the iterative configuration of internal organ states to maintain stability under stress. AO utilizes numerical... Read more
Egret Swarm Optimization Algorithm: An Evolutionary Computation Approach for Model Free Optimization
Key finding: Introduces the Egret Swarm Optimization Algorithm (ESOA), which simulates hunting behaviors of two egret species using a learnable sit-and-wait strategy (exploitation) and aggressive random wandering (exploration). This dual... Read more
Key finding: Presents Red Panda Optimization (RPO), a parameter-free bio-inspired metaheuristic modeling red pandas’ foraging and tree-climbing behaviors for exploration and exploitation phases respectively. RPO’s mathematical formulation... Read more
3. How can clustering techniques be integrated with evolutionary algorithms to effectively identify and exploit multiple local and global optima in complex multimodal optimization landscapes?
This research area addresses the challenge of multimodality in optimization problems, where multiple local and global minima coexist. Integrating clustering operators within evolutionary algorithms enables partitioning the population into meaningful clusters that correspond to distinct basins of attraction around minima. Such integration supports simultaneous discovery and refinement of multiple solutions, improves convergence speed, and enhances robustness by confining search efforts adaptively within promising regions. The use of clustering algorithms like k-windows within EAs represents an advancement toward efficient multi-minima optimization.
Key finding: Proposes a novel clustering operator incorporating the unsupervised k-windows algorithm within Differential Evolution (DE) to utilize accumulated population data for identifying separate regions containing clusters of... Read more
Key finding: Develops hybrid evolutionary algorithms combining GA and PSO to solve nonlinear constrained problems. The hybridization allows leveraging PSO’s social learning for local exploitation within clusters of solutions, while GA... Read more
Key finding: Utilizes a hybrid genetic algorithm integrating GA and PSO to manage multi-objective optimization by facilitating simultaneous exploration of different trade-off clusters in the solution space. The method leverages clustering... Read more