UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

A Hybrid Object-Oriented Environment Integrating Neural Networks and Expert Systems

Khebbal, Sukhdev Singh; (1995) A Hybrid Object-Oriented Environment Integrating Neural Networks and Expert Systems. Doctoral thesis (Ph.D), UCL (University College London). Green open access

[thumbnail of A_hybrid_object-oriented_envir.pdf]

Download (12MB) | Preview


This thesis investigates the fundamental issues and methods for building hybrid systems. It presents an object-oriented hybrid environment (ORBERON) for integrating symbolic processing techniques, such as expert systems, and adaptive processing techniques, such as neural networks. The ORBERON environment was the inspiration for the European Community Esprit III HANSA project, which has produced a framework for the rapid integration of industry standard tools and novel artificial intelligence shells. The design and implementation of the HANSA framework is also detailed in this thesis. This thesis is composed of five major parts: (i) the fundamental issues, arguments and methodologies for integrating symbolic and adaptive processing; (ii) the design and implementation of the ORBERON environment; (iii) the implementation of a complex real-world problem using the ORBERON environment; (iv) the design of the Esprit HANSA Framework and (v) the assessment of the hybrid approach, the ORBERON environment and the hybrid solution to the real-world problem. To establish the arguments for integrating symbolic and adaptive processing, the structures, applications, strengths and weaknesses of expert systems and neural networks are examined. These techniques were chosen because they represent the most popular and well developed examples from the symbolic and adaptive disciplines. To capitalise on their complementary properties, various methods for integrating these techniques are reviewed and a development cycle for constructing hybrid systems is described. To support this hybrid approach, a classification scheme is proposed that categorises hybrid systems with regards to functionality, communication, and processing architecture. The ORBERON environment is constructed around an object-oriented framework, so that each processing technique can be represented as an object and can communicate with other techniques via a message-passing mechanism. This approach allows design, communication and execution autonomy for techniques. The techniques inherit a standard communication protocol when encapsulated in the environment's Generic Interface Object. Techniques active within the environment are managed via a Dynamic Environment Manager. The design and implementation of these core components is examined in detail. The functional validation of the ORBERON environment was undertaken by implementing solutions to a small proof of concept problem for Profit Trend Analysis and a complex real-world Cargo Consignment problem, from British Airways. The design and implementation of hybrid solutions to both problems are detailed. The hybrid solution to the Cargo Consignment problem and the viability of adopting a hybrid philosophy, was assessed by comparing the results of the hybrid approach with the results of the existing purely expert system approach. The ORBERON environment was assessed with reference to current object-oriented integration methods. The main research contributions of this work have been: (i) the analysis and comparison of symbolic and adaptive techniques; (ii) the formulation of a novel classification scheme for hybrid systems; (iii) the design of a development cycle for hybrid systems; and (iv) the ORBERON environment for building hybrid systems. This research has also produced an edited book titled "Intelligent Hybrid Systems" and produced five publications.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: A Hybrid Object-Oriented Environment Integrating Neural Networks and Expert Systems
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Thesis digitised by ProQuest.
URI: https://discovery.ucl.ac.uk/id/eprint/10103799
Downloads since deposit
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item