Towards extending adaptive self-organizing concurrent system architecture

This paper discusses an adaptive self-organizing concurrent system (ASOCS), whose functionality relies on incremental, supervised learning paradigm. ASOCS can be trained to recognize categories in response to an arbitrary binary input vector. ASOCS is comprised of many boolean processing nodes distributed throughout the system. An adaptation unit is connected to all the logic nodes in order to supervise consistency checking and minimize the system function representation. Depending upon the adaptation unit directives, boolean processing nodes interactively pass messages, add new nodes, delete redundant nodes from the network. These actions lead to self-modification and self-organization. After presenting the pertinent features of a generic ASOCS, this paper discusses an extension leading to improved generalization and more compact knowledge representation