Environment for Knowledge Acquisition by Agents in Internet Infrastructure Restricted Domains

Abstract

The development environment of this project is Internet Infrastructure, where theso-called routing domains are represented by Autonomous Systems (ASes). Theresearch creates an environment that provides resources and facilities to allowautonomy to agents hosted and distributed by ASes. This environment containsresources for agents to collaborate with each other and with humans using techniquesassociated to a data structure known as blockchain. The environment provides aknowledge base that stores explicit knowledge and mechanisms that infer implicitknowledge. Explicit knowledge is obtained from multi-modal and unstructureddatabases, available outside the local environment and appropriately transformedinto databases called corpus. Such corpus undergoes transformation processes thatinvolve natural language processing techniques, including summarization, and forma set of intermediate databases, prepared to be used in an orderly manner tocapture knowledge to be used to build explicit knowledge into a knowledge base.Another testing environment is proposed, to exercise algorithms that consolidate theproposed ideas. Intermediate bases are constructed and proposed to favor machinelearning algorithms that will add knowledge to the corresponding base. All thecomponents of the proposed environment act together and/or separately to providea mechanism for feedback and preservation of the knowledge acquired, keeping itpersistent and following the paradigm of learning that never ends. The knowledgebase and the machine learning results feed a multi-agent model divided into fourlayers. The text shows that such model was used for the creation and maintenanceof objects of a server very well known on the Internet called IRR, avoiding humanintervention and creating reliability, so far absent in this server. The success ofthe new implementation of the IRR was one of the motivations for the creation,expansion and improvement of an environment suitable for autonomous agents.

A Machine Learning Ontology

“…The example given above, associated with the Internet Infrastructure, was made the choice to include the machine learning applications over the Autonomous Systems sub-domain, name given to the computer networks that represent, together, the Internet [10]. Thus, it was a matter of doing a literature review (non-systematic) to survey the primary studies that could produce the data and qualify them.…”

“…The other items that appear in the image of this figure are components that complete the SKAU environment that, although important are outside the scope of this paper. Adapted from [9] or [10]. developed in Java and its source code is released under a free license (the Mozilla Public License).…”