Clustering Concepts into Higher-Level Entities using Neural Network-like Structures

摘要

Previous work has described linking mechanisms and how they might be used in a cognitive model that could even begin to think [1][2][3]. One key problem is enabling the system to autonomously form its own concept structures from the information that is presented. This is particularly difficult if the information is unstructured, for example, individual concept values being presented in unstructured groups. This paper suggests an addition to the current model that would allow it to filter the unstructured information to form higher-level concept chains that would represent something in the real world. The new architecture also starts to resemble a traditional feedforward neural network, suggesting what future directions the research might take.