Artificial Searching Swarm Algorithm for Black-Box Optimization

Summary

The article describes the Artificial Searching Swarm Algorithm (ASSA), a population-based method for optimizing difficult, potentially discontinuous or mixed-type objective functions. Each agent can move toward a recently improved agent in response to a short-lived signal, search using its personal and the swarm’s best-known positions, or take a random step when those references provide little direction. A shared record preserves the global best, while each agent retains its own best position. The method normalizes distances across differently scaled variables and includes boundary handling and support for discrete search spaces.

The author implements ASSA in a unified test framework and reports experiments on standard optimization benchmarks, including Hilly, Forest, Megacity, Rosenbrock, and Griewank. The article references earlier comparisons in which ASSA performed well on high-dimensional problems, but the supplied excerpt omits much of the current test results, so specific comparative conclusions cannot be assessed here. The stated advantage is a small parameter set; stagnation is identified as a weakness. Suggested improvements include diversification and changes to how signals or moves are handled.

Key ideas

• ASSA uses short-lived improvement signals to direct agents toward promising regions.
• Agents also search using their personal best and the swarm’s global best, with random movement when these references offer little guidance.
• Normalized distances help make movement consistent across variables with different ranges.
• The article evaluates the algorithm on benchmark functions, but the provided text does not include enough current results to verify detailed performance claims.
• The method is presented as having few parameters, with stagnation listed as a limitation.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.