📄 http:^^anytime.cs.umass.edu
字号:
Date: Wed, 20 Nov 1996 19:47:31 GMTServer: NCSA/1.4.2Content-type: text/html<html><head><title>Resource-Bounded Reasoning -- Home Page</title></head><body bgcolor="#FFFFFF" text="#000000" link="#0440DD" vlink="#0440DD" background="pix/br.gif"><table><tr><td valign=top><!WA0><A HREF="http://anytime.cs.umass.edu/index.html"><!WA1><img src="http://anytime.cs.umass.edu/pix/clock.gif" align=topwidth="177" height="143" border="2"></a><p></td><td valign=top> </td><td width=350 valign=top><br><P><h2>The Resource-Bounded Reasoning Research Group</h2><!WA2><a href="http://www.cs.umass.edu">Department of Computer Science</a><br><!WA3><a href="http://www.umass.edu">University of Massachusetts</a><br>Box 34610, Lederle Graduate Research Center <br>Amherst, MA 01003-4610 <br>voice: +1-413-545-4189 fax: +1-413-545-1249<br> <br><P><br><P> </td></tr></table><table><tr><td width=125 valign=top><B> <!WA4><a href="http://anytime.cs.umass.edu/personnel.html">Personnel</a><font size="+2"> </font><br> <!WA5><a href="http://anytime.cs.umass.edu/research.html">Research</a><font size="+2"> </font><br> <!WA6><a href="http://anytime.cs.umass.edu/publications.html">Publications</a><font size="+2"> </font><br> <!WA7><a href="http://anytime.cs.umass.edu/symposia.html">Symposia</a><font size="+2"> </font><br> <!WA8><a href="http://anytime.cs.umass.edu/talks.html">Talks</a><font size="+2"> </font><br> <!WA9><a href="http://anytime.cs.umass.edu/affiliations.html">Affiliations</a><font size="+2"> </font><br></B></td><td valign=top> </td> <td width=400 valign=top>The Resource-Bounded Reasoning Research Group is part of the<!WA10><a href="http://www.cs.umass.edu/">Department of Computer Science</a> atthe <!WA11><a href="http://www.umass.edu/"> University of Massachusetts,Amherst</a>.The group, headed by Professor <!WA12><a href="http://anytime.cs.umass.edu/~shlomo">Shlomo Zilberstein</a>,studies the implications of limited computational resources on thedesign of intelligent agents. The group conducts research indecision theory, real-time planning, autonomous agentarchitectures and reasoning under uncertainty.<h2>What is resource-bounded reasoning?</h2>Resource-bounded reasoning is an emerging field within artificialintelligence that is concerned with the construction of intelligentsystems that can operate in real-time environments under uncertainty andlimited computational resources. Research in this field covers the construction, composition andmeta-level control of computational methods that allow small quantitiesof computational commodities -- such as time, memory, or information -- tobe traded for gains in the value of computed results.<h2>Why is it needed?</h2>The need to employ resource-bounded reasoning techniques is based on asimple, but general, observation. In many complex domains, thecomputational resources required to reach an optimal decision reduce theoverall utility of the result. This observation covers a wide range ofapplications such as medical diagnosis and treatment, combinatorialoptimization, probabilistic inference, mobile robot navigation, andinformation gathering. What is common to all these problems is thatit is not feasible (computationally) or desirable (economically) tocompute the optimal answer. Moreover, taking the cost of decision-makinginto account is not an easy task, since the "optimal" level ofdeliberation varies from situation to situation. It is therefore beneficial to build systems that can tradeoff computational resources for quality of results.<h2>Advantages of resource-bounded reasoning</h2>From the early days of AI, heuristic methods and "satisficing" techniqueshave been used to address the problem of computational complexity.Resource-bounded reasoning techniques have two important advantages over those previous approaches: they shift the attention from design-time solutions to more flexible, run-time solutions, and they seek to optimize rather than satisfice solution quality.<P>The shift to run-time control of deliberation improves the capability ofintelligent systems to deal with two primary sources of uncertainty.The first source is internal to the system and relates to itscapability to produce incrementally improving solutions and to assesstheir quality.The second source of uncertainty is external and relates tounpredictable change in the environment in which the system operates.Run-time control of deliberation seeks to reduce the effectof these uncertainties on the performance of the system.<P>Optimization of decision quality is another distinctive feature of resource-bounded reasoning. That is, instead of building systems that find a "good" answer, the goal of resource-bounded reasoning techniquesis to find an "optimal" answer. Optimality, however, is defined withrespect to the system knowledge and computational capabilities. Typically,an optimal answer does not require maximal solution quality. Hence thesesystems are sometimes referred to as "bounded optimal" or"bounded rational".<hr noshade><address><!WA13><A HREF="mailto:shlomo@cs.umass.edu">shlomo@cs.umass.edu</A></address></td></tr></table></body> </html>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -