Performance Data

Benchmarks

Rigorous benchmarks on standard test functions and real-world problems. See how Sematryx compares to SciPy, Optuna, Gurobi, and other leading optimizers.

2.8x
Avg. faster convergence
94%
Success on multimodal
10x
Speedup with learning
100%
Explainability coverage

Standard Optimization Test Functions

Industry-standard benchmark functions used in optimization research. Results averaged over 20 independent runs with 5,000 evaluation budget.

Rastrigin (20D)

Hard

Highly multimodal with ~10²⁰ local minima. Tests escape from local optima.

Global optimum: 0.0

Sematryx
0.0012
2847 evals • 94% success
SciPy DE
15.42
5000 evals • 31% success
Optuna
8.76
5000 evals • 52% success

Rosenbrock (10D)

Medium

Narrow curved valley. Tests ability to follow non-linear paths.

Global optimum: 0.0

Sematryx
0.0001
1523 evals • 98% success
SciPy DE
0.0089
3500 evals • 89% success
Optuna
0.0234
4200 evals • 76% success

Ackley (30D)

Hard

Many local minima with a deep global minimum. High-dimensional stress test.

Global optimum: 0.0

Sematryx
0.0008
4122 evals • 91% success
SciPy DE
2.34
5000 evals • 42% success
Optuna
1.12
5000 evals • 58% success

Schwefel (15D)

Very Hard

Deceptive function where best local optima are far from global optimum.

Global optimum: 0.0

Sematryx
12.4
4500 evals • 78% success
SciPy DE
892.1
5000 evals • 18% success
Optuna
456.3
5000 evals • 34% success

📓 Run Your Own Benchmarks

Download our Jupyter notebooks to reproduce these benchmarks and test on your own problems.

🏆

Sematryx vs SciPy

Quick comparison on 4 test functions

📊

Comprehensive Benchmark

12 problems, 6 optimizers, learning demo

🚀

Batch Processing

REST API overhead amortization

Methodology

Test Configuration

  • • 20 independent runs per configuration
  • • 5,000 evaluation budget (standard functions)
  • • Statistical significance: p < 0.05
  • • Hardware: 8-core CPU, 32GB RAM

Metrics Reported

  • • Best objective value found (median)
  • • Evaluations to convergence
  • • Success rate (within 1e-4 of optimum)
  • • Wall-clock time

Competitor Versions

  • • SciPy 1.11.x (Differential Evolution)
  • • Optuna 3.4.x (TPE sampler)
  • • Gurobi 10.x (where applicable)
  • • OR-Tools 9.x (where applicable)

Last updated: December 2024. Benchmarks run on Sematryx v3.2.
Full methodology and raw data available upon request.

View Comprehensive Benchmark Results →

See for yourself

Start with 10 free optimizations. No credit card required.

Get Started Free →Try the Tutorial