Evolving Artificial Neural Networks

Evolving Artificial Neural Networks (EANNs) are the hallmark synthetic version of emergent intelligence. Though they vary along several dimensions, they often exhibit emergence across all three primary adaptive levels: evolution, development, and learning. They are also the default representation in evolutionary robotics (ER), the field that most tangibly illustrates the true potential of biologically inspired routes to AI. In their seminal ER textbook Nolfi and Floreano (2000) cite several reasons for the popularity of ANNs and EANNs for robotic control: There is normally a smooth mapping between ANN parameters and the behaviors they produce. This enhances adaptive search. It is