The ability to remember past experiences enables a system to improve its performance as well as its competence. For example, a system might be able solve problems faster by adapting previous solutions. Additional tasks, such as avoiding unwanted behavior by detecting potential problems, monitoring long-term goals by remembering what subgoals have been achieved, and reflection on past actions, become feasible. As the tasks that an intelligent system accomplishes become more and more complex, so does the experience it acquires in the process. Such experience has a temporal extent and is expressed in terms of concepts and relations with deep semantics associated to them. Memory systems should be able to deal with the temporal aspect of experience, exploit this semantic knowledge for storage and retrieval and do so in a scalable fashion. However, relying just on experience will not achieve a broad coverage, as it needs to be used in conjunction with other reasoning mechanisms. That is why we need the ability to add episodic memory functionality to intelligent systems. Today's knowledge-based systems are complex software applications and the ability to develop them in a modular fashion, using generic, reusable components is essential. We propose to separate the episodic memory from the system that uses it and to build a generic, reusable memory module that can be attached to a variety of applications in order to provide this functionality. Its goal is to provide accurate, scalable, efficient and content-addressable access to prior episodes. Having such a reusable memory module should allow research to focus on the generic aspects of memory representation, organization and retrieval and its interaction with the external application and it should also reduce the complexity of the overall system. In this dissertation we propose a set of general requirements that any memory module should provide regarding memory encoding, storage and retrieval. We present an implementation that satisfies these requirements and evaluate it on three different tasks: plan synthesis, plan recognition and Physics problem solving. The memory module proved easily adaptable to these tasks, providing fast, accurate and scalable retrieval.
| Identifer | oai:union.ndltd.org:UTEXAS/oai:repositories.lib.utexas.edu:2152/3326 |
| Date | 28 August 2008 |
| Creators | Tecuci, Dan Gabriel |
| Contributors | Porter, Bruce, 1956- |
| Source Sets | University of Texas |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | electronic |
| Rights | Copyright © is held by the author. Presentation of this material on the Libraries' web site by University Libraries, The University of Texas at Austin was made possible under a limited license grant from the author who has retained all copyrights in the works. |
Page generated in 0.0021 seconds