Deterministic Oscillatory Search (DOS)
DOS is the founding four’s outlier: a fully deterministic population-based optimizer that
“combines the advantages of traditional gradient-based methods with the efficiency of swarm
algorithms” — no randomness. Source: andrey-dik‘s MQL5 implementation (C_AO_DOS,
S_DOS_Velocity).
How it works
Particles move via three mechanisms:
- Oscillatory movement — on detecting fitness deterioration a particle “bounces” (reverses direction, halves velocity), a zigzag that localizes nearby extrema.
- Fitness-slope tracking — each particle holds a three-state flag (improving / worsening / unknown) to judge movement quality.
- Swarming — when oscillation stops improving, particles move toward the global best.
Distinctive feature
Determinism / reproducibility: “given the same initial conditions, the algorithm always arrives at the same result.” The natural experiment on whether stochasticity is load-bearing in metaheuristics (see exploration-vs-exploitation and synthesis).
Benchmark (context-relative)
Ranked 45th (32.36%) on the population-optimization-benchmark — underperformed stochastic methods, “particularly on low-dimensional functions” — suggesting the randomness it removes was doing useful work. Benchmark-relative (no-free-lunch-theorem).
Related
metaheuristic-optimization · exploration-vs-exploitation · population-optimization-benchmark · no-free-lunch-theorem · andrey-dik · exchange-market-algorithm