gomea 1.0.6

Library for the use of various variants of the Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA).

README

The Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) is a state-of-the-art Model-Based Evolutionary Algorithm (MBEA) with an ongoing line of research into various domains of (evolutionary) optimization. The initial release of the GOMEA library (v1.0.0) is described in the publication:

• Bouter, A., & Bosman, P.A.N. (2023). A Joint Python/C++ Library for Efficient yet Accessible Black-Box and Gray-Box Optimization with GOMEA. In Proceedings of the Genetic and Evolutionary Computation Conference Companion

The current release of the GOMEA library supports optimization in the following domains:

• Discrete single-objective optimization
• Real-valued single-objective optimization

Support for multi-objective optimization in both of these domains is planned for a future release.

Relevant most recent publications are the following:

• Dushatskiy, A., Virgolin, M., Bouter, A., Thierens, D., & Bosman, P.A.N. (2021). Parameterless Gene-pool Optimal Mixing Evolutionary Algorithms. arXiv preprint arXiv:2109.05259.
• Bouter, A., Alderliesten, T., & Bosman, P.A.N. (2021). Achieving highly scalable evolutionary real-valued optimization by exploiting partial evaluations. Evolutionary computation, 29(1), 129-155.

INSTALL

The most straightforward installation option is through pip, as follows.

pip install gomea

To build from source, it is recommended to create a conda virtual environment with the dependencies listed in environment.yml, as follows:

conda env create --file=environment.yml
conda activate gomea

Then compile and install the code by executing the following in the root GOMEA folder.

make install

You can now start using the library by calling import gomea in your code, or by running any of the demo scripts present in the demos directory.

RUN

Running a simple benchmark problem, e.g., the Rosenbrock function, can be done as follows:

import gomea
frv = gomea.fitness.RosenbrockFunction(10,value_to_reach=1e-10)
lm = gomea.linkage.Univariate()
rvgom = gomea.RealValuedGOMEA(fitness=frv, linkage_model=lm, lower_init_range=-115, upper_init_range=-100)
result = rvgom.run()

Plotting a simple convergence plot can then be done using:

import matplotlib.pyplot as plt
plt.grid()
plt.yscale('log')
plt.xscale('log')
plt.xlabel('Number of evaluations')
plt.ylabel('Best objective value')
plt.plot(result['evaluations'],result['best_obj_val'])

Rather than using a predefined fitness function, it is also possible to define your own custom fitness function for Black-Box Optimization (BBO) or Gray-Box Optimization (GBO). The definition of a custom (real-valued) BBO function can be done as follows.

class CustomRosenbrockFunction(gomea.fitness.BBOFitnessFunctionRealValued):
    def objective_function( self, objective_index, variables ):
        f = 0
        for i in range(len(variables)-1):
            x = variables[i]
            y = variables[i+1]
            f += 100*(y-x*x)*(y-x*x) + (1.0-x)*(1.0-x)
        return f

dim = 10
f_rosenbrock = CustomRosenbrockFunction(dim,value_to_reach=1e-10)

The definition of a GBO function requires defining the input and output of each subfunction. Thorough definitions of this are given in the EvoSOFT workshop paper. An example of the Rosenbrock function defined in a GBO setting is as follows:

class CustomRosenbrockFunction(gomea.fitness.GBOFitnessFunctionRealValued):
    def number_of_subfunctions( self ):
        return self.number_of_variables-1
    
    def inputs_to_subfunction( self, subfunction_index ):
        return [subfunction_index, subfunction_index+1]

    def subfunction(self, subfunction_index, variables):
        x = variables[subfunction_index]
        y = variables[subfunction_index+1]
        return 100*(y-x*x)*(y-x*x) + (1.0-x)*(1.0-x)
        
dim = 10
f_rosenbrock = CustomRosenbrockFunction(dim,value_to_reach=1e-10)

Further examples are given in the demos directory.