Spiral Swarm DSS

Deterministic optimization with 100% reproducible results. Same input, same output, every time. Up to 78% fewer function evaluations on multimodal problems.

Request Trial View Documentation

100%

Reproducible results

78%

Fewer evaluations

Hyperparameters to tune

10⁻¹¹

Precision achieved

import thalosforge as tf

# Expensive CFD simulation
def cfd_simulation(params):
    # Each call costs $50 in compute
    return run_cfd(params)

bounds = [(0, 1)] * 12

# DSS: deterministic, evaluation-efficient
result = tf.optimize(
    cfd_simulation,
    bounds=bounds,
    method='dss',
    maxevals=200
)

# Run again → identical result
result2 = tf.optimize(...)
assert result.x == result2.x  # Always true

Why DSS

When reproducibility and efficiency matter more than anything

100% Deterministic

Zero randomness. Fibonacci lattice sampling with geometric contraction guarantees identical results every run. No seed management, no variance, no surprises.

Zero Hyperparameters

Works out of the box. No learning rates, population sizes, or mutation rates to tune. Just pass your function and bounds—DSS handles the rest.

Evaluation Efficient

Up to 78% fewer function calls on multimodal problems compared to PSO and GA. When each evaluation costs real money, this translates directly to savings.

β-Adaptive Contraction

Automatically balances exploration and exploitation. Contracts faster when stagnating, slower when improving—without any manual intervention.

Regulatory Ready

Deterministic results are auditable and defensible. Perfect for FDA submissions, aerospace certifications, and financial compliance where you must explain your optimization.

Geometrically Interpretable

The search path is fully traceable. You can visualize exactly how DSS explored your solution space—no black-box mystery.

Stochastic Methods vs DSS

Stochastic Optimizers (PSO, GA, DE)

Different result every run
5-10 hyperparameters to tune
Cannot reproduce exact solution
Requires multiple runs for confidence
Hard to explain to regulators
Random seed management overhead

Deterministic Spiral Search (DSS)

Identical result every run
Zero hyperparameters required
Exact reproducibility guaranteed
Single run is definitive
Fully auditable search path
No seed management needed

Use Cases

Where determinism and efficiency are non-negotiable

CFD/FEA Optimization

Each simulation costs $50-$500 in compute time. DSS's 78% evaluation reduction means real cost savings. Deterministic results satisfy engineering review requirements.

Pharmaceutical Development

FDA requires documentation of optimization decisions. DSS provides a fully traceable, reproducible search path that auditors can verify independently.

Aerospace Design

Safety-critical systems require deterministic optimization. DSS delivers identical wing profiles, control parameters, or structural configurations every time.

Financial Model Calibration

Calibrate complex models with reproducible results. When regulators ask how you arrived at your risk parameters, you can show them exactly.