Memetic Algorithm with CMA-ES using LS Chains
---------------------------------------------

This is a software created in C++ by Daniel Molina Cabrera daniel.molina@uca.es proposed in

D. Molina, M. Lozano, C. García-Martínez, F. Herrera, Memetic Algorithms for Continuous Optimization Based on Local Search Chains. 
Evolutionary Computation, 18:1 (2010) 27-63, doi: 10.1162/evco.2010.18.1.18102

How to get it
-------------

The last stable version are available from the research group of the author: Soft Computing and Intelligent Information Systems
http://sci2s.ugr.es/EAMHCO/ 

Brieftly description of the algorithm
-------------------------------------

This algorithm uses a Memetic Algorithm Continuous Optimisation (MACO) model, MA with LS Chains.
This model employs the concept of {\it LS chain} to adjust the LS intensity, assigning to each 
individual an LS intensity that depends on its features, by chaining different LS applications. In that model, an individual
resulting from an LS invocation may later become the initial point of a subsequent LS application, adopting the final strategy 
parameter values achieved by the former as its initial ones. In this way, the continuous LS method may adaptively fit its 
strategy parameters to the particular features of the search zones, increasing the LS effort over the most promising 
solutions and regions. 

MA-CMA-Chains is composed by:
    
    * An Steady-State Genetic Algorithm (SSGA). The SSGA applied was specifically designed to promote high population 
diversity levels by means of the combination of the BLX-alpha crossover operator with a high value for its associated 
parameter (alpha=0.5), and an adequated combination of selective method and replacement strategy. It uses
the Negative Assortative Mating (NAM) as its selection method, and Replacement Worst (RW) as its replacement strategy.

    * CMA-ES algorithm as its Local Search method. CMA-ES was proposed in 

    Hansen N, Müller SD, Koumoutsakos P (2003). Reducing the time complexity of the derandomized evolution strategy with 
    covariance matrix adaptation (CMA-ES). Evolutionary Computation, 11(1) pp. 1–18, 2003.

    * The MA with LS Chains model proposed, using an initial LS step size of 500 evaluations, investing the 50% of
    fitness evaluations in the SSGA and the other half of number evaluations in the LS. 

See the first reference to obtain a complete detailed description. 

Install guide
-------------
It is required for install the software:

* gcc/g++. It has been tested with gcc version 4.4.1, but .

* CMake version 2.4 or upper (it has been tested with version 2.6). 

* Makefile

CMake is a building software (like Makefile), and it has to be installed in conjuntion with make to compile 
the library. 

* Library realea (from http://sci2s.ugr.es/EAMHCO) was installed. MA-CMA-Chains and Libreal must be decompressed
from the same directory. Thus, it must have the following structure:

/anydirectory
    /macmachains
    /realea

Steps:
------

* Decompress the file:
tar zxvf macmachains.tgz

* Install the library (if it was not previously installed)
tar zxvf realea_lib.tgz
cd realea; cmake .; make install; cd -

* Create makefiles
cmake .

* Compile 
make 

Program
-------
There is a program to show the algorithm. 

* main_malschains_cec2005, it uses MA-LS-Chains for solve CEC2005 test suite. Note: it requires the directory input_data
to run.

The syntax is the following:

usage: ./macmachains_cec2005 fun dim [maxrun=25]
Where	fun is the number of function (1 to 25)
	dim is the dimensionality (10|30|50)
	maxrun is the maximum times (default=25)