E-Net is a new distributed evolutionary learning system that evolves neural-network-based pattern recognition systems (PRSs) with limited human interaction. This system orchestrates a multiplicity of evolutionary and classical learning techniques to synthesize feature detectors, select sets of cooperative features, and assemble classifiers. Feature detectors are represented as feed-forward neural networks and recognition systems are defined using a collection of networks. E-Net evolves network topologies and trains weights to form accurate recognition systems using a computationally efficient process that gradually extends primitive network topologies to form increasingly discriminating structures. The evolutionary search process effectively explores the space of candidate topologies by manipulating populations of feature detectors and recognition systems using variation operators such as crossover and mutation. The majority of evolutionary learning techniques have been designed to perform parameter optimization. E-Net is designed to perform both synthesis and optimization. Consequently, many novel concepts and techniques are introduced in this research that expedite the gradual synthesis of structure, such as the new multitiered selection process used in E-Net's evolutionary algorithm that avoids premature convergence to complex topological structures.

The paper introduces Multi-Tiered Tournament Selection. Traditional tournament selection algorithms are appropriate for single objective optimization problems but are too limited for the multi-objective task of evolving complete recognition systems. Recognition systems need to be accurate as well as small to improve generalization performance. Multi-tiered Tournament Selection is shown to improve search for smaller neural network recognition systems.

A new assembly line inspection system that permits an operator to teach the system what is to be considered good and bad without any need for computer reprogramming is developed and demonstrated. The feasibility of using neural networks combined with a simple feature extraction algorithm to make visual inspection systems which learn is demonstrated. The demonstration system can separate round parts in the class of problems which have all of the required information in a circular band concentric to the center of the part and which have visually detectable features. The machine is shown to have good parts and flawed parts. In the latter case, the type of flaw is entered in the computer. Preprocessing is used to provide position and rotation invariance. A feedforward network is then trained to provide the correct output. The system is shown to perform reliably.<>