Deterministic Oscillatory Search for Reproducible Optimization
Summary
The article describes Deterministic Oscillatory Search, a population-based metaheuristic for optimizing multidimensional objective functions without random numbers. Particles begin at systematically distributed positions and track whether movement improves or worsens fitness. When a particle's progress reverses, it changes direction and reduces its velocity, producing oscillations intended to narrow in on an extremum. If local movement no longer helps, a swarming step steers it toward the best solution found by the population.
The article outlines the algorithm's state tracking, movement rules, boundary handling, and implementation, then discusses comparisons on test functions. It presents reproducibility as a benefit: fixed starting conditions lead to repeatable results. The author reports speed and simplicity as strengths, while noting variation on low-dimensional functions. The evidence is limited to the optimization tests described; performance on synthetic objective functions does not establish an advantage for trading strategies or financial data. The article also cautions that some implementations of canonical algorithms were modified, so comparisons and conclusions depend on the test setup.
Key ideas
- DOS replaces random particle initialization and movement with deterministic rules.
- Particles track fitness slope and reverse with reduced velocity when their progress worsens.
- A swarming mechanism moves particles toward the best solution found when local exploration stalls.
- Repeatable runs support reproducibility, but low-dimensional tests show result scatter.
- Optimization benchmark results do not establish trading performance.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.