Open Access. Powered by Scholars. Published by Universities.®
Articles 1 - 3 of 3
Full-Text Articles in Entire DC Network
Deriving Optimal Solutions From Incomplete Knowledge Bases, Shawn A. Northrop
Deriving Optimal Solutions From Incomplete Knowledge Bases, Shawn A. Northrop
Theses and Dissertations
Many real world domains cannot be represented using Bayesian Networks due to the need for complete probability tables and acyclic knowledge. However, Bayesian Knowledge Bases (BKBs) are a viable method for representing these incomplete domains, but very little research has been performed on inferencing with them. This paper presents three inference engines for extracting optimal solutions from three distinct BKB subclasses: singly- connected, multiply-connected with mutually exclusive cycles, and cyclic. The singly-connected inference engine has a worst case polynomial run time. Performance improvement techniques for increasing inference engine speed are discussed, in addition to a new tool for measuring incompleteness …
Probabilistic Knowledge Base Validation, Howard T. Gleason
Probabilistic Knowledge Base Validation, Howard T. Gleason
Theses and Dissertations
Our work develops a new methodology and tool for the validation of probabilistic knowledge bases throughout their lifecycle. The methodology minimizes user interaction by automatically modifying incorrect knowledge; only the occurrence of incomplete knowledge involves interaction. These gains are realized by combining and modifying techniques borrowed from rule-based and artificial neural network validation strategies. The presented methodology is demonstrated through BVAL, which is designed for a new knowledge representation, the Bayesian Knowledge Base. This knowledge representation accommodates incomplete knowledge while remaining firmly grounded in probability theory.
Semantic Interpretation Of An Artificial Neural Network, Stanley D. Kinderknecht
Semantic Interpretation Of An Artificial Neural Network, Stanley D. Kinderknecht
Theses and Dissertations
Recent advances in machine learning theory have opened the door for applications to many difficult problem domains. One area that has achieved great success for stock market analysis/prediction is artificial neural networks. However, knowledge embedded in the neural network is not easily translated into symbolic form. Recent research, exploring the viability of merging artificial neural networks with traditional rule-based expert systems, has achieved limited success. In particular, extracting production (IF.. THEN) rules from a trained neural net based on connection weights provides a valid set of rules only when neuron outputs are close to 0 or 1 (e.g. the output …