Multi Expression Programming library

version 2024.04.18.0-beta

Implements the Multi Expression Programming (MEP) technique for solving symbolic regression, classification (binary and multiclass), and time-series prediction problems.

MEP is a Genetic Programming (GP) variant with a linear representation of chromosomes.

MEP introduced a unique feature: the ability to encode multiple solutions in the same chromosome. This means that we can explore much more from the search space compared to other techniques that encode a single solution in the chromosome. In most cases, this advantage comes with no penalty regarding running time or resources involved.

To compile:

C++ 11 is required due to the use of C++ 11 threads.

Create a new project and add:

• all files from src and
• one file from main folder (which contains the main function).

Include paths must point to the include folder of this project.

If you use the MS compiler, add _CRT_SECURE_NO_WARNINGS and _CRT_NONSTDC_NO_DEPRECATE to the preprocessor definitions.

To run:

You need some a file with training data. We provided several files (located in the data folder) for test:

• bulding1.csv for symbolic regression problems,
• cancer1.csv for binary classification problems with 0/1 output.
• cancer1_output1-1.csv for binary classification problems with -1/1 output.
• iris.txt for multiclass classification problems.
• fibonacci.txt for univariate time-series.
• wage_growth.csv for multi-variate time-series.

Make sure that the instruction (from the main function):

if (!training_data->from_tabular_file("../data/building1.csv")) ...

has the correct path of the file.

Graphical user interface

libmep is used by MEPX.

Documentation

https://github.com/mepx/libmep/wiki

Documentation is currently obsoleted. It will be updated soon.

More info:

mepx.org

mepx.github.io

https://github.com/mepx

Discussion Group:

https://groups.google.com/d/forum/mepx

Contact author:

mihai.oltean@gmail.com