Enhancing Scalability in Genetic Programming With Adaptable Constraints Type Constraints and Automatically Defined Functions

Why this code exists

Code for CGPF2.1 and ACGPF2.1. Modified original lilgp1.02 -> lilgp1.03 kernel to handle bigger problems in support of 2019 dissertation for completion of Ph.D. in mathematical and computational sciences at the University of Missouri - St. Louis (UMSL).

Why lilgp1.02 was changed

lilgp1.02 population statistics were all saved as ints. Integer overflows happened on large gp runs. So, population statistics were changed from int to long

Software needed to run this

Operating System Ubuntu 16.04 or better. Not tested for Mac. Won't run on Windows.

gcc any modern version should work -- compiler for lilgp

GNU parallel version 20161222 or better -- helps to push lilgp runs to as many cores as possible

bash -- lots of bash scripts that get the job done

sqlite3 3.28.0 -- holds stats for runs, fed into R

R scripting version 3.4.2 (2017-09-28) -- stats processing

autogen (GNU AutoGen) 5.18.7 -- help in creating consistent code for different gp problems

lush (lisp) http://lush.sourceforge.net/ -- some text file wrangling

To build all code for running problems

Steps:

1. cd into lilgp1.03/app
2. ./runprob.bash
3. Go get a cup of coffee.... it will take about 3 days to a week or more depending on your machine.

To run code for a particular problem

cd into a subdirectory named starting with prefix prob
./start.bash -- will run everything for a particular problem

Key Subdirectories

app -- holds experiments

app/mg52 -- holds results of data for 52 generations

app/mg104 -- holds results of data for 104 generations

app/mg156 -- holds results of data for 156 generations

app/mg208 -- holds results of data for 208 generations

Other information

Warning -- this will fill up your hard disks with lots of data. Well over 1TB.