Adaptive pattern recognition in a real-world environment

Pattern recognition systems.;Neural networks (Computer science)

Bairaktaris, Dmitrios ; Livesey, Mike ; Livesey, Mike

University:University of St Andrews

Department:Computer Science (School of)

关键词: Pattern recognition systems.;    Neural networks (Computer science);   

Others  :  https://research-repository.st-andrews.ac.uk/bitstream/handle/10023/9261/DimitriosBairaktarisPhDThesis.pdf?sequence=2&isAllowed=y

来源: DR-NTU

【 摘 要 】

This thesis introduces and explores the notion of a real-world environmentwith respect to adaptive pattern recognition and neural network systems. Itthen examines the individual properties of a real-world environment andproposes Continuous Adaptation, Persistence of information and Context-sensitiverecognition to be the major design criteria a neural network system ina real-world environment should satisfy.Based on these criteria, it then assesses the performance of Hopfield networksand Associative Memory systems and identifies their operational limitations.This leads to the introduction of Randomized Internal Representations, a novelclass of neural network systems which stores information in a fullydistributed way yet is capable of encoding and utilizing context.It then assesses the performance of Competitive Learning and AdaptiveResonance Theory systems and again having identified their operationalweakness, it describes the Dynamic Adaptation Scheme which satisfies allthree design criteria for a real-world environment.