Archery Algorithm: Population-Based Optimization and Benchmark Tests

Summary

The document explains a population-based stochastic optimizer inspired by archers aiming at a target. Each candidate solution is an agent whose coordinates are updated using another agent selected according to fitness-based probabilities. The described implementation adds a chance to copy a selected agent’s coordinates, Gaussian randomness, a training-intensity parameter, bounds on candidate positions, and memory of previous positions. The algorithm retains the best solution found until the iteration limit is reached.

The article reports tests on standard objective functions and compares the modified algorithm with other optimizers. It characterizes the tested version as fast, self-adaptive, and broadly convergent, while noting a tendency to become stuck on some low-dimensional functions. These findings come from the author’s experiments; the document cautions that implementations may differ from canonical algorithms and that its conclusions depend on those experiments. The method is presented as a general optimization technique, with finance and trading mentioned as possible application areas rather than evaluated use cases.

Key ideas

• Agents represent candidate solutions and update their positions by learning from fitness-weighted selections of other agents.
• The described implementation combines copying, Gaussian perturbations, coordinate bounds, and a randomized training-intensity parameter.
• Agents retain previous positions and fitness values, while the algorithm tracks the best solution across iterations.
• Benchmark experiments describe balanced convergence across dimensions but identify possible stagnation on low-dimensional functions.
• The reported results assess optimization functions rather than trading performance.

Tags

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