Integrations

Surfaces integrates seamlessly with popular optimization frameworks. Use your favorite optimizer with any Surfaces test function.

---

Supported Frameworks

scipy

Scientific Python’s optimization module. Built-in to_scipy() conversion.

scipy

Optuna

Hyperparameter optimization framework. Works with suggest_* API.

Optuna

SMAC

Sequential Model-based Algorithm Configuration. ConfigSpace integration.

SMAC

Ray Tune

Distributed hyperparameter tuning. Scalable experimentation.

Ray Tune

Gradient-Free-Optimizers

Simple gradient-free optimization. Native search space compatibility.

Gradient-Free-Optimizers

Hyperactive

Advanced optimization toolkit. Direct integration.

Hyperactive

---

Quick Comparison

Framework	Type	Scale	Best For
scipy	Classical	Single machine	Quick prototyping, gradient methods
Optuna	Bayesian	Distributed	HPO, pruning, visualization
SMAC	Bayesian	Single machine	Algorithm configuration, research
Ray Tune	Various	Distributed	Large-scale, multi-GPU
GFO	Gradient-free	Single machine	Simple API, exploration
Hyperactive	Meta-heuristic	Single machine	Advanced search strategies

---

Common Pattern

All integrations follow a similar pattern:

1. Create a Surfaces test function

2. Extract the search space

3. Define an objective that calls the test function

4. Run the optimizer

from surfaces.test_functions.algebraic import RastriginFunction

# 1. Create function
func = RastriginFunction(n_dim=5)

# 2. Get search space
space = func.search_space()

# 3. Define objective
def objective(params):
    return func(params)

# 4. Run optimizer (framework-specific)
# ...

---

Choosing a Framework

For quick experiments:

Use scipy or GFO. Minimal setup, fast iteration.

For serious benchmarking:

Use Optuna or SMAC. Better optimization, logging, visualization.

For large-scale studies:

Use Ray Tune. Distributed execution, checkpointing, scalability.

For research:

Use SMAC or custom. Fine-grained control, reproducibility.

---