📄 http:^^www.cs.utexas.edu^users^bert^research.html
字号:
MIME-Version: 1.0
Server: CERN/3.0
Date: Monday, 06-Jan-97 22:48:31 GMT
Content-Type: text/html
Content-Length: 3278
Last-Modified: Wednesday, 30-Oct-96 21:53:06 GMT
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"><HTML><HEAD> <TITLE>Dissertation Research Overview</TITLE> <META NAME="GENERATOR" CONTENT="Mozilla/3.0Gold (X11; I; SunOS 5.5 sun4c) [Netscape]"></HEAD><BODY><H1>Dissertation Research Overview</H1><H2>THIS PAGE IS CURRENTLY UNDER CONSTRUCTION!!! </H2><P>The overall theme of my research is reasoning with and refining imprecisemodels of physical processes. Because we cannot know the world precisely,imprecision is implicit in any modeling task. However, it is often thecase that imprecision is simply ignored since it has no place in many existingmodeling and simulation methodologies. Sometimes, this is of no consequence,but in other cases, the mismatch between prediction and true behavior maylead to erroneous conclusions. In fields where imprecision is addressed(for example, in robust control), the methods are often resticted in termsof the types of imprecision allowed or are not guaranteed to produce allthe predictions inherent in the model or produce predictions that are tooweak to provide useful guidance. What is needed is a general method forrepresenting and reasoning with imprecison that can capture the typicaltypes of imprecision inherent in modeling tasks and can produce predictionsthat maintain precision consistent with the model imprecision. In the <!WA0><A HREF="http://www.cs.utexas.edu/users/bert/sqsim.html">nextsection</A>, I describe SQSim, a representation and simulation method thatmeets these goals. </P><P>While imprecision is inherent in any modeling task, it is also truethat a model's precision can be improved with experience with the underlyingphysical process. Learning a model from a combination of prior knowledgeand empirical data is the task of system identification. Typically, identificationmethods work under the assumption that a parametric model of the processexists and searches the model space for the precise model that best matchesthe empirical data. Confidence bounds on the parameters (and hence theprediction) can also be determined to represent the uncertainty inherentin using a finite amount of data. This method can be quite efficient incases where the search space has good properties (such as an easily computablegradient). For cases where this does not hold, however, (for example, ifwe allow functional as well as parametric uncertainty), finding an optimalmodel may be very difficult. In addition, identification methods requirethat empirical data be informative enough to provide complete experiencewith the system over the desired operating range. If such conditions cannotbe met, the resulting model may differ greatly from the true model. Inthe <!WA1><A HREF="http://www.cs.utexas.edu/users/bert/squid.html">third section</A>,I describe SQUID, a method for refining an existing model by using possiblyuninformative data. </P><P>SQSim and SQUID provide two key technologies for automating the model-buildingtask. By combining them with a method for postulating initial qualitativemodels from data, one could construct a system that can automatically constructmodels using a combination of empirical and prior knowledge.</P><P><!WA2><A HREF="http://www.cs.utexas.edu/users/bert/index.html">BKay </A></P></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 + -