This is a C# project that illustrates how Genetic Algorithms (GA) can be used for agent shaping. In this case the agents are immaginary faces. Using GA it is possible to find solutions to optimization problems. The algorithm recombines the population of possible solutions in order to find the best one. The candidate solution must be represented as an array of bits called chromosome. At the beginning a random population of agents is generated, in this first step the agents look very different.

In the second step a function called fitness is used in order to select the best individuals. In evoface there is not a fitness function because it is the user that select the best individual of the population. After the selection some kind of recombination operators can be applied to the single agent or to a group of agents.

When you select a specific face the software take the agent's chromosome and mutate it. The resulting offsprings are a recombination of the original agent. When a mutation happens it switches a single bit inside the chromosome and it influences the appearence of the offspring. If you select an agent the remaining windows will be fill with 8 offsprings

The mutation probability can be changed (the default value is 1.0%), higher the value higher the chromosome variability. For example, if you choose the same agent but you set an higher mutation probability (10.0%) the resulting offsprings will be more different than the precedent ones.

Evoface is a very basic example of GA and can be used as an introductory code that can be expanded in many ways.

There are two main classes in evoface: Chromosome and Agent. The Chromosome class is a generic and reusable class that store a list of bytes for representing a single solution to the problem. A set of methods is part of the class and permits to mutate the genotype and clone it. The mutation method in particular, permits to access each bit of the genotype and switch it with a given probability. This kind of mutation can generate a new genotype that can be slightly or higly different from the original one. The Agent class uses the Chromosome to encapsulate the features associated to the agent, it permits to translate the chromosome to a face (genotype to phenotype). Each byte inside the chromosome represent a particular feature, unrolling the chromosome and analyzing each single byte permits to have a complete description of the agent.

The repository provides a ready to use executable and the complete source code. The solution file can be opened with Visual Studio 2015, I did not tested it with older versions of VS. The project is self-contained and does not require any external library.

1. Clone the repository or download it here
2. Open the file evoface.sln with Visual Studio.
3. Compile the code and enjoy.

The repository provides also a pre-compiled version of the software in the folder evoface/bin/Release. The pre-compiled executable can be started also in Linux using Wine.