Collection

The collection module provides a unified interface to browse, filter, and select test functions. It includes pre-defined suites for common benchmarking scenarios.

---

Why Use Collection?

Instead of manually selecting functions:

# Manual selection - tedious
from surfaces.test_functions.algebraic import (
    SphereFunction,
    RastriginFunction,
    RosenbrockFunction,
    AckleyFunction,
    # ... many more
)

functions = [
    SphereFunction(n_dim=10),
    RastriginFunction(n_dim=10),
    # ...
]

Use the collection:

# With collection - simple
from surfaces import collection

functions = collection.standard.instantiate(n_dim=10)

---

Pre-defined Suites

Quick Suite

Fast sanity check (5 functions) for smoke tests and development iteration.

from surfaces import collection

functions = collection.quick.instantiate(n_dim=10)

for func in functions:
    print(f"{func.__class__.__name__}")

Includes: Sphere, Ackley, Rosenbrock, Rastrigin, Griewank

Standard Suite

Academic comparison (15 functions) covering diverse landscape types.

from surfaces import collection

functions = collection.standard.instantiate(n_dim=10)

Includes: Classic functions for publication-quality benchmarking.

BBOB Suite

Full COCO/BBOB benchmark (24 functions) for standardized comparison.

from surfaces import collection

functions = collection.bbob.instantiate(n_dim=10)

CEC Suites

IEEE CEC competition benchmarks.

from surfaces import collection

# CEC 2014 (30 functions)
functions_2014 = collection.cec2014.instantiate(n_dim=10)

# CEC 2017 (10 functions)
functions_2017 = collection.cec2017.instantiate(n_dim=10)

Engineering Suite

Constrained engineering design problems (5 functions).

from surfaces import collection

functions = collection.engineering.instantiate()

for func in functions:
    print(f"{func.__class__.__name__}: {len(func.constraints)} constraints")

---

Filtering Functions

Filter functions by properties:

from surfaces import collection

# All unimodal functions
unimodal = collection.filter(unimodal=True)

# Convex algebraic functions
convex = collection.filter(category="algebraic", convex=True)

# Separable functions
separable = collection.filter(separable=True)

Available filter criteria:

• category: “algebraic”, “bbob”, “cec”, “engineering”, “ml”

• unimodal: True/False

• convex: True/False

• separable: True/False

• scalable: True/False

• n_dim: Specific dimension or None for variable

---

Searching Functions

Search by name or tagline:

from surfaces import collection

# Find all Rastrigin variants
rastrigin = collection.search("rastrigin")
print(f"Found {len(rastrigin)} Rastrigin functions")

for func_cls in rastrigin:
    print(f"  - {func_cls.__name__}")

---

Set Operations

Combine collections using set operations:

from surfaces import collection

# Union: all functions from both suites
combined = collection.quick + collection.engineering

# Intersection: functions in both
common = collection.filter(scalable=True) & collection.bbob

# Difference: remove engineering from all
non_engineering = collection - collection.engineering

---

Using with Optimizers

from surfaces import collection
from gradient_free_optimizers import BayesianOptimizer

functions = collection.standard.instantiate(n_dim=10)
results = {}

for func in functions:
    opt = BayesianOptimizer(func.search_space)
    opt.search(func, n_iter=100)

    results[func.__class__.__name__] = opt.best_score

# Print results table
for name, score in sorted(results.items(), key=lambda x: x[1]):
    print(f"{name:25s}: {score:.6f}")

---

Listing Available Suites

from surfaces import collection

for name, count in collection.list_suites().items():
    print(f"{name}: {count} functions")

---

Iterating Over All Functions

The collection itself contains all available functions:

from surfaces import collection

print(f"Total functions: {len(collection)}")

# Iterate over all
for func_cls in collection:
    print(func_cls.__name__)

---

Next Steps

• Test Functions - All available test functions

• Integrations - Use collection with optimization frameworks

• /api_reference/collection - Complete collection API