The EPlex group specializes in several key areas:

• Hypercube-based Neuroevolution of Augmenting Topologies (HyperNEAT): HyperNEAT is a method for evolving neural networks with large numbers of nodes and connections that was created here in the EPlex group at UCF.  It uses connective CPPNs to generate regular connectivity patterns that can be produced at scalable resolutions.  HyperNEAT in effect  turns the original NEAT into a special kind of indirect encoding for large-scale neural networks.  It exploits the fact that connectivity patterns are isomorphic to spatial patterns in a higher-dimensional space.
   
• Complexification: Complexification is the process through which new genes are added to the genome over generations, thereby gradually increasing the dimensionality of the search space, and hence the potential complexity of the phenotypes.  Complexification is a means to find solutions in high-dimensional spaces even if it is not practical to start searching directly in such spaces.
  
  
• Indirect Encoding, Developmental Encoding, and Gene Reuse: If genes can be reused in the mapping between genotype and phenotype, then the genotype space can contain far fewer dimensions than the phenotype space, making the search for high-level complexity tractable.  For example, the 30,000 genes in the human genome produce a brain with 100 trillion connections through a process of development.
  
  
• Compositional Pattern Producing Networks (CPPNs): CPPNs are a novel abstraction of biological development developed here in the EPlex group.
  
  
  
  
• NeuroEvolution of Augmenting Topologies (NEAT): The NEAT method (created by EPlex director Kenneth Stanley) is the first principled implementation of complexifying evolution.  We utilize the general principles of NEAT as an underlying methodology for our evolutionary experiments.
  

• Artificial Neural Networks (ANNs): Our long-term goal is to evolve extremely complex ANNs.  Through complexification and indirect encoding an intermediate goal is to evolve useful ANNs with tens of thousands of neurons or more.
  
  
• Competitive Coevolution: In sophisticated competitive domains in which it is difficult to formalize a fitness function, the only way to spark a chain of increasing evolutionary complexity may be through a direct competition among evolving individuals.
  
  
• Video games, simulation, and rtNEAT: The real-time NEAT algorithm can be used to evolve intelligent non-player-characters (NPCs) and novel content in video games and training applications in real time.  As new methods for evolving complexity are developed, they can be applied to video games and simulation by integrating them with rtNEAT in the future.  The picture below shows the NERO video game, which uses rtNEAT to allow players to train robots in real time to perform tasks.
  
  
  
  
• Group Behavior: In natural evolving populations there is often group interaction during evolutionary evaluation (i.e. during life).  This interaction often leads to complex role assignment and cooperation.