Hybrid AI and Machine Learning Systems

Many different Artificial Intelligence and Machine Learning architectures and algorithms have been developed, each with their own strengths and weaknesses that make them particularly suited to certain classes of problems. Two approaches to artificial intelligence are dealt with here: subsumption architecture, and neural networks. S.A.’s are good at building up complex behaviors from sets of simpler ones, when a problem can be broken down into independent peices. Neural networks are good at memorizing associations and making inferences about new data based on stored memories, but can take impractically long to train when the datasets are large. Performance on various tasks can be improved by combining the two approaches to take advantage of the strengths of both.