libfgen: Main Page

Introduction

libfgen is a library that implements an efficient and customizable genetic algorithm (GA). It also provides particle swarm optimization (PSO) functionality and an interface for real-valued function minimization or model fitting. It is written in C with a partial C++ wrapper.

fgen

fgen is the interface to the genetic algorithm implementation. In addition to bitstring operators, permutation-based operators are also provided and custom seeding, mutation and crossover operators can be defined by the application, also allowing use of real-valued parameters. Archipelagos of multiple concurrent GAs are supported. Steady-state evolution is also supported. Internal threading allows efficient parallel computation. The library is thread-safe.

fgenpp

fgenpp is the C++ wrapper interface to fgen (the genetic algorithm implementation). It offers direct equivalents to the functions in the C version. A sub-class must be used to define the application specific operators.

fpso

fpso is the interface to the particle swarm optimization implementation, which provides a simple PSO with GBEST or LBEST topology and other customizable parameters. Although it can converge quickly for simple problems, for more complex higher dimensional problems the genetic algorithm generally performs much better.

ffit

ffit is the interface to straightforward multi-dimensional function minimization. It can be used for model fitting where a set of floating point parameters is evaluated resulting in an error value. It is also convenient for general optimization problems with real-valued parameters. A genetic algorithm or PSO can be used to find the global minimum. Parameters can be distributed within a range with a profile other than linear (for example logarithmic or binomial) to optimize the searching process. When using the genetic algorithm ffit uses 16, 32 or 64-bit bitstring representations mapped to real values, or alternatively normalized real values.

Examples

ones.c is a trivial GA problem using bitstrings.
func.c is a trivial function maximization problem using a GA.
tsp.c implements the Travelling Salesman Problem using a genetic algorithm with permutation representation and specialized mutation and crossover operators.
tsp_archipelago.c does the same with an archipelago of concurrent genetic algorithms. It also supports steady-state evolution.
tsp_archipelago_cpp.cpp is a C++ version of the latter using the C++ wrapper API.
pi_pso.c is a trivial example of PSO.
ffit_rastrigin.c uses the ffit API to find the minimum of a multi-dimensional Rastrigin function using a genetic algorithm or PSO.
ffit_rosenbrock.c does the same for the two-dimensional Rosenbrock-function.
rastrigin_double.c is an example of real-valued chromosomes with custom operators applied to the Rastrigin function.
pict is graphical example using ffit which decomposes an image into filled rectangles. It also supports texture compression to ETC1 or DXT1 format. It requires GTK+ 3.
tsp is a graphical example of the Travelling Salesman Problem. It requires GTK+ 3.
gp is an example program that implements linear genetic programming.

libfgen can be downloaded from the sourceforge.net page for the project: http://sourceforge.net/projects/libfgen/.