Computational issues of nonmonotonic reasoning /
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1993.
|
| Subjects: | |
| Online Access: | ProQuest, Abstract Link to OAKTrust copy |
| Abstract: | From the earliest days of the modern computer age, humankind has dreamed of machines that use "common sense" to reason about the world. However, common sense as displayed by humans largely consists of techniques for dealing with the enormous amount of information provided by the five senses. Since this information can overwhelm even the human brain, the brain can only process it effectively in the form of generalizations and "rules of thumb." Hence, much of its information is not complete; it only captures the most general cases. So, despite efforts to provide vast and thorough knowledge bases, it has become clear that reasoning machines must have the ability to model their worlds in the face of incomplete information about those worlds. Towards this end, many types of "nonmonotonic" reasoning techniques have been proposed. Using these techniques a computer can "change its mind" about what it believes to be true in the light of new information. In recent years, nonmonotonic reasoning has grown to be one of the most widely researched fields in computer science. Unfortunately, while it is apparent that this type of reasoning is very complex, little research has focussed on just how complex it truly is. More importantly, little is known about what kinds of restricted forms of nonmonotonic reasoning can be practically implemented on computers. The research presented herein is aimed at answering the question, "Are there computationally prohibitive characteristics that are part of the nature of nonmonotonic reasoning?" Perhaps a clearer way to state this is, "Can any useful form of nonmonotonic reasoning be automated in an efficient manner?" The general approach taken in this research is as follows. First, a semantic model of generalized nonmonotonic reasoning that lends itself to computational analysis is developed. Then, a methodology for using this model to analyze restricted forms of nonmonotonic reasoning is discussed. Some restricted classes of nonmonotonic reasoning systems are then defined and analyzed. Finally, the correctness and practical importance of this work is demonstrated by using it to analyze some previous work; in particular, several classes of "model-preference" default logic systems are analyzed. |
|---|---|
| Item Description: | Vita. "Major subject: Computer Science." |
| Physical Description: | x, 146 leaves : illustrations ; 28 cm |
| Bibliography: | Includes bibliographical references. |