Fractal-Based Algorithm for Continuous Optimization

Summary

The article explains the Fractal-Based Algorithm (FBA), a metaheuristic for continuous optimization that allocates more search effort to areas containing high-quality candidate solutions. It partitions the search space into subspaces, ranks those areas by how many promising points they contain, subdivides selected regions, and generates new candidates there. A mutation step sends some points into less explored areas, while population integration preserves strong solutions. The described implementation exposes parameters for population size, promising-point and subspace shares, mutation, and the number of intervals per dimension.

The article presents pseudocode and an MQL5 class structure, then reports comparisons on standard test functions. It characterizes FBA as stable on medium- and high-dimensional functions, while noting greater result variation on low-dimensional ones. The article gives no detailed numeric results in the supplied text, and its assessment is limited to its test setup. It also cautions that the implementation modifies canonical algorithms, so the description and experimental conclusions should be treated as specific to this implementation. FBA is a general optimization method, not a trading strategy or evidence of market performance.

Key ideas

• FBA partitions the search space and repeatedly focuses on subspaces containing many high-quality candidates.
• Promising regions are subdivided to support finer local search.
• Mutation sends a fraction of candidates into less predictable areas to maintain exploration.
• The implementation balances global exploration and local refinement through configurable population and selection parameters.
• The reported tests describe stable behavior in medium and high dimensions but more variable results in low dimensions.

Tags

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