Performance analysis of a cognitive analogical reasoner

Renewed interest in cognitive information processing has identified the need for novel computing architectures that can efficiently and dynamically support reasoning and learning with emphasis on memory structures. A large number of cognitive processing methods have been implemented in software on conventional computing architectures. This paper describes detailed execution profiling of a particular cognitive method for analogical reasoning called SAGE™. SAGE™ is a symbolic-connectionist, performance oriented analogical reasoner. The primary contribution of the work is to identify the deficiencies and bottlenecks of SAGE™ to inform the development of improved architectures for cognitive information processing, and especially for analogical reasoning.