metaheuristic-designer 0.1.2

A python package to desing metaheuristic optimization algorithms from components.

Metaheuristic-designer

Python implementation of a general framework for the design and execution of metaheuristic algorithms.

Created with the purpose of creating the necesary tools for the analysis and design of metaheuristics in a more granular way, choosing algorithm components and adding them to a general search framework.

Inspired by this article: Swan, Jerry, et al. "Metaheuristics “in the large”." European Journal of Operational Research 297.2 (2022): 393-406.

Mostly following the book: Eiben, Agoston E., and James E. Smith. Introduction to evolutionary computing. Springer-Verlag Berlin Heidelberg, 2015.

Instalation

Since this package is not available in PyPi at the moment, you will have to install it manually, preferably in a virtual environment.

To do this, get the package from the latest release, in this case, the file you are looking for is "metaheuristic_designer-0.1.0-py3-none-any.whl". and install it with the following command:

pip install .

Examples

• There are 2 scripts to test this repository:
  • "examples/exec_basic.py": Optimize a simple function, in this case, the "sphere" function that calculates the squared norm of a vector, we want a vector that minmizes this function. There are two possible flags that can be added:
    • "-a [Alg]" use one of the available algorithms, the choices are:
      • HillClimb: simple hill climbing algorithm.
      • LocalSearch: take the best of 20 randomly chosen neighbours.
      • ES: (100+150)-ES, basic evolutionary strategy.
      • HS: Harmony search algorithm.
      • GA: genetic algorithm.
      • SA: simulated annealing algorithm.
      • DE: DE/best/1, differential evolution algorithm.
      • PSO: simple particle swarm algorithm.
      • NoSearch: no search is done.
    • "-m" use a memetic search like structure, do local search after mutation.
  • "examples/exec_basic.py": Evolve an image so that it matches the one given as an input. Recieves mostly the same parameters except for one for indicating the input image:
    • "-i [Image path]" read the image and evolve a random image into this one.

It is recomended that you create a virtual environment to test the examples. This is done with the following commands:

python -m venv venv
source venv/bin/activate

Once you have activate the virtual environment, you can execute one of the examples like this:

python examples/example_basic.py

To run the tests you need to install nox, to execute the tests use the command

nox test

General parameters

• Stopping conditions:
  • stop_cond: stopping condition, there are various options
    • "neval": stop after a given number of evaluations of the fitness function
    • "ngen": stop after a given number of generations
    • "time": stop after a fixed amount of execution time (real time, not CPU time)
    • "fit_target": stop after reaching a desired fitness, accounting for whether we have maximization and minimization
    • All of the above can be combined with the logical operators 'or' or 'and' to make more complex stopping conditions, an example can be "ngen or time" which will stop when the number of generations or the time limit is reached.
  • Neval: number of evaluations of the fitness function
  • Ngen: number of generations
  • time_limit: execution time limit given in seconds
  • fit_target: value of the fitness function we want to reach
• Display options:
  • verbose: shows a report of the state of the algorithm periodicaly
  • v_timer: amount of time between each report

Operators available

• 1 point cross (1point)
• 2 point cross (2point)
• Multipoint cross (Multipoint)
• Weighted average cross (WeightedAvg)
• BLXalpha cross (BLXalpha)
• SBX cross (SBX)
• Multi-individual cross (Multicross)
• Xor with random vector (Xor)
• Cross two vectors with Xor (XorCross)
• Permute of vector components (Perm)
• Random mutation of vector components (MutRand/MutNoise)
• Add noise following a prob. distribution (RandNoise)
• Approximate population with a prob. distribution and sample on some vector components (MutSample)
• Approximate population with a prob. distribution and sample (RandSample)
• Mutation adding Gaussian noise (Gauss)
• Mutation adding Cauchy noise (Cauchy)
• Mutation adding Laplace noise (Laplace)
• Mutation adding Uniform noise (Uniform)
• Differential evolution operators (DE/best/1, DE/best/2, DE/rand/1, DE/rand/2, DE/current-to-rand/1, DE/current-to-best/1, DE/current-to-pbest/1)
• Particle swarm algorithm step (PSO)
• Firefly algorithm step (Firefly)
• Place fixed vector into the population (Dummy) [only intended for debugging]
• Generate a completely random solution (Random)
• Replace some of the components with random values (RandomMask)
• Apply a custom function (Custom)
• Return a copy of the individual (Nothing)

Parent selection methods available

• Tournament (Tournament)
• Taking the n best parents (Best)
• Roulette method (Roulette)
• Stochastic Universal Sampling (SUS)
• Take all parents (Nothing)

Survivor selection methods avaliable

• Elitism (Elitism)
• Conditional Elitism (CondElitism)
• Generational (Generational)
• Compare each parent with its child (One-to-one)
• Compare each parent with its child, take the child with a probability (Prob-one-to-one)
• (λ+μ) like method ( (m+n) )
• (λ,μ) like method ( (m,n) )
• Selection method in the Coral Reef optimization algorithm (CRO)