Establishing basis for learning algorithms
Abstract
The CONNECT Integrated Project aims at enabling continuous composition of networked systems, by developing techniques for synthesizing connectors. A prerequisite for synthesis is to learn about the interaction behavior of networked peers. The role of WP4 is to develop techniques for learning models of networked peers and middleware through exploratory interaction. In this deliverable, we survey the CONNECT process in order to derive requirements for the participating learning techniques. We also report on a number of case studies from which such requirements are extracted. These include requirements on the ability to interact with the networked peer, whose behavior is being learned, as well as requirements on the learned model, in order that it can be used for subsequent manipulation in the CONNECT process, such as connector synthesis. A major challenge is to extract and maintain detailed information about the interface of the networked peer, as well as to relate this information to the produced model, in which abstractions have been employed in order to make synthesis tractable. We describe our approach for representing and maintaining this information, and how we have adopted existing learning techniques to make use of it. Ontologies are proposed as a suitable vehicle for representing the information. We also report on work performed to develop and adapt existing learning tools, in order that they be suitably useful in the CONNECT process. This deliverable summarizes the progress and achievements during Year 1 in WP4.
Domains
Software Engineering [cs.SE]
Origin : Files produced by the author(s)
Loading...