Another Numeric optimization and Metaheuristics Library

A library to do your metaheuristics and numeric combinatorial stuff.

To install use:

pip install anmetal

See /test folder for examples of use.

Content

Numeric optimization

Iterative optimization functions (one solution)

• Euler method
• Newton method

Metaheuristics

Real input (Population-based)

• Artificial Bee Colony (ABC)
• Ant Colony Optimization (ACO)
• Artificial Fish Swarm Algorithm (AFSA)
• Bat Algorithm
• Blackhole Algorithm
• Cuckoo Search
• Firefly Algorithm
• Harmony Search (HS)
• Particle Swarm Optimization (PSO)
• Particle Swarm Optimization with Leap
• Greedy
• Greedy with Leap

Categorical input

• Genetic Algorithm
• Genetic Algorithm with Leap

Problems and gold-standard functions

NP-hard problems

• Real problems

  • Partition problem
  • Subset problem

• Categorical problems

  • Knapsack
  • Sudoku (without initial matrix, just random)

Non linear functions

• One input (1-D)

  • F1 (https://doi.org/10.1007/s00521-017-3088-3)
  • F3 (https://doi.org/10.1007/s00521-017-3088-3)

• Two inputs (2-D)

  • Camelback (https://doi.org/10.1007/s00521-017-3088-3)
  • Goldsteinprice (https://doi.org/10.1007/s00521-017-3088-3)
  • Pshubert1 (https://doi.org/10.1007/s00521-017-3088-3)
  • Pshubert2 (https://doi.org/10.1007/s00521-017-3088-3)
  • Shubert (https://doi.org/10.1007/s00521-017-3088-3)
  • Quartic (https://doi.org/10.1007/s00521-017-3088-3)

• N inputs (N-D)

  • Brown1 (https://doi.org/10.1007/s00521-017-3088-3)
  • Brown3 (https://doi.org/10.1007/s00521-017-3088-3)
  • F10n (https://doi.org/10.1007/s00521-017-3088-3)
  • F15n (https://doi.org/10.1007/s00521-017-3088-3)
  • Sphere (https://doi.org/10.1007/s00521-018-3512-3)
  • Rosenbrock (https://doi.org/10.1007/s00521-018-3512-3)
  • Griewank (https://doi.org/10.1007/s00521-018-3512-3)
  • Rastrigrin (https://doi.org/10.1007/s00521-018-3512-3)
  • Sumsquares (https://doi.org/10.1007/s00521-018-3512-3)
  • Michalewicz (https://doi.org/10.1007/s00521-018-3512-3)
  • Quartic (https://doi.org/10.1007/s00521-018-3512-3)
  • Schwefel (https://doi.org/10.1007/s00521-018-3512-3)
  • Penalty (https://doi.org/10.1007/s00521-018-3512-3)

Additional Features

Binarization functions

• sShape1
• sShape2
• sShape3
• sShape4
• vShape1
• vShape2
• vShape3
• vShape4
• erf

Binarization strategies

• standard
• complement
• static_probability
• elitist

Example Usage

See the /test folder for complete examples. Here's a quick overview of running different metaheuristics:

# Example with Partition Problem
from anmetal.problems.nphard_real import Partition_Real
from anmetal.population.PSO.PSOMH_Real import PSOMH_Real

# Create problem instance
problem = Partition_Real(seed=0, num_dims=200)

# Create and run metaheuristic
mh = PSOMH_Real(problem.min_x, problem.max_x, problem.ndim, False,
                problem.objective_function, problem.repair_function,
                problem.preprocess_function)

# Run optimization
fitness, solution = mh.run(verbose=True, iterations=100, population=30,
                         omega=0.8, phi_g=1, phi_p=0.5, seed=115)

Algorithm Parameters

Each metaheuristic has its own set of parameters. Here are some common ones:

• Common Parameters

  • iterations: Number of iterations
  • population: Population size
  • seed: Random seed for reproducibility
  • verbose: Whether to print progress

• Algorithm-Specific Parameters

  • ABC: limit
  • ACO: evaporation_rate, alpha, beta
  • BAT: fmin, fmax, A, r0
  • CUCKOO: pa
  • FIREFLY: alpha, beta0, gamma
  • GA: mutation_rate, crossover_rate
  • HS: hmcr, par, bw
  • PSO: omega, phi_g, phi_p

For detailed parameter descriptions and recommended values, see the respective algorithm implementations in the source code.