I am currently working with cellular automata and evolutionary
computations to solve the Majority Classification
Problem***. I hope to beat the existing record of about 85%
completion in the MCP. I am working with test arrays that are of length 145.
The rules are based off of a 7-neighbor system and therefore need to be at
least of length
128 (there are 128 possible combinations of 7 binary cells). However, I
also added an extra cell where the value (success rate)
of the rule is stored.
I do not know what kind of genetic algorithm I will use.
I will most likely have to do a lot of testing and trial-and-error to
figure out which variable values work the best.
If I complete the stated problem in a timely fashion,
there are many direction that I would like to expand. Instead of
working with a 1-D cellular automata, I could expand the program to
deal with 2-D or perhaps even 3-D landscapes. Also, I might try
comparing the results of simultaneous evolution of the whole system
verses random evolution of isolated cells. There is also the
possibility of dealing with systems that are not just black and white.
Lastly, if I complete the MCP and wish to try something else, I can
apply my cellular automata system to a variety of other problems.
***Majority Classification Problem: The goal of the
Majority Classification Problem is to use evolutionary computations to
evolve sets of rules that will manipulate a cellular automata system
towards its majority. For example, if there is an array with 65% black
cells, the rule (a defined array) should eventually develop the
landscape to an all-black array.