GPJIT is a project mainly written in C, it's free.
Genetic programming with Just-In-Time compilation of expression trees.
GPJIT is a simple genetic programming software that tries to create an algebraic expression as an approximation to an unknown R -> R function. It''s innovative in the sense that it generates x86 assembly instructions in order to improve the speed of fitness evaluation of each individual in the population.
GPJIT was developed as an assignment for the course "Introduction to Genetic Algorithms in Robotics" at University of São Paulo, circa 2009.
The expression trees are represented in a traditional way using the ETTree and Element structs. Each branch of the tree may contain an arithmetic operator or a function with their parameters/operands in the child nodes. Each leaf of the tree contains either a constant value or a reference to a variable.
The Assembler object traverses an expression tree in postorder DFS, outputting x86 assembly directly - this is possible because the x86 FPU is stack-based.
The fitness function is simply evaluated as the sum of the absolute errors, i.e. $abs(freal(xi) - fapprox(xi))$ over an interval (as defined in ga.c: evalFitness).
GPJIT generates, mutates and applies cross-over operators to the expression tree in order to improve the population fitness, as suggested by Koza.
The assembly code generated was created by hand using nasm and, therefore, will only work in the platform in which it was developed: x86 Linux 32bits. The generated code may be ported to a 64 bit platform by changing the code to be generated in the assembly.c file. A better and more generic solution could be developed by using a platform-agnostic JIT library, such as libjit.
Due to the simplistic approach to code generation, it was a design decision to limit the generated functions to those that could be evaluated using only the FPU stack of an x86 machine, without resorting to local variables. Due to the limited space of the FPU stack, this implies that the maximum tree depth that can be handled by the program is 6.
GPJIT depends on the allegro graphics library to render an on-screen visualization of the function being approximated VS the best individual of the population.
In a debian-based distribution:
$ sudo apt-get install liballegro4.2-dev*$ cd src
$ make
$ ./gpjitThe MIT License
Copyright (c) 2011 Tiago Camolesi
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.