Rule Weights Research Articles

A probabilistic rule-based expert system

This paper explores a medical expert system combining techniques of Bayesian network modelling with ideas of weighted inference rules. The weights of the individual rules can be estimated objectively from a training set of actual cases; and they can be used in a Monte Carlo simulation to estimate objectively conditional probabilities of diagnosis given particular combinations of symptoms. The paper describes and evaluates a medical expert system built according to this design. The diagnostic accuracy of the program was found to be similar to that obtained through the usual application of Bayes theorem with the assumption of conditional independence of symptoms given disease, even though the Bayesian classifier has more than 70 times as many numerical parameters. The method may be promising in cases where small training sets do not permit accurate estimation of large numbers of parameters.

Adaptive fuzzy systems

Conventional fuzzy control systems using PID (proportional-integral-derivative) control and their limitations are discussed. Ways to incorporate adaptivity are examined. The functioning of adaptive fuzzy logic and adaptive fuzzy control systems is described. The use of rule weights is explained. >

On the stability of the classical collocation method for Cauchy-type singular integral equations

The classical collocation method for Cauchy-type singular integral equations of the second kind on an interval using Gaussian quadrature rules with respect to Jacobi nodes is investigated. In general, the nodes and the weights of the quadrature rules can only be evaluated approximately. In the present paper the question is answered how precise our knowledge about them must be in order to ensure that the corresponding algebraic system of equations is solvable yet and its solution is close to the solution of the exact system.

Three kinds of case-based learning in sheet metal manufacturing

The paper outlines three applications of case-based learning in sheet metal manufacturing by utilization of artificial intelligence, each based on different concepts. One topic is the synthesis with feature processing for the process outline on a CNC bending machine. Statistics on fail and success influence weights of rules in an expert system. Another example is the classification of sheet metal parts. Real objects and their components are represented by a frame like data structure, which is embedded in a semantic network. Grouping of work pieces is done by a decision method of Bayes. This method is also used for part diagnosis to avoid possible defects of tools and machines.

The Use of MIDST, an Expert System Building Tool for Solving Geological Problems

Many geological problems can be solved using knowledge-based approaches, but system design is slowed by complex interactions that occur between knowledge engineers and geological experts. MIDST, a user-friendly set of tools, partially circumvents knowledge engineers and allows expert geologists to build expert systems directly. These tools address: (1) knowledge acquisition, and (2) knowledge refinement and updating. MIDST is a general purpose tool, and its application is demonstrated with the development of a system for characterizing fields, plays, and prospects (PLAYMAKER). Knowledge acquisition is driven by a graphic outliner which creates a conceptual hierarchical structure of nodes that defines the problem-solving domain (in this case field characterization). The user defines relevant geological classes and concepts (e.g., reservoir qualities, porosity and permeability, depositional setting, and facies), and provides relationships between nodal concepts as directed inferential links (e.g., depositional setting is linked to facies, and so to porosity and permeability). Next, a rule-editor is used to formulate rules that define links. Inferencing requires starting with low-level node characters in the conceptual structure (e.g., basin setting), and using results from these characters to successively infer characteristics at higher level nodes (e.g., porosity and permeability). Nodes can be created and removed, and links changed asmore » rules are developed. The knowledge refinement tool enables expert geologists to evaluate the effectiveness of expert systems by identifying deficiencies in rules when consultation results are unsatisfactory. This can be attributed to either incomplete rules, inconsistency between rule pairs, or lack of fine tuning in rule weights.« less

Fundamentals of Fuzzy Prolog

In this paper, a Fuzzy Prolog is introduced that is based on the fuzzy resolution principle. There are three levels of concern: (1) the fuzzy truth values of predicates that depend on several values of individual variables, (2) the weights of rules that depend on several fuzzy truth values of premise and conclusion, and (3) the confidence of resolvent that depends on the fuzzy resolution principle. With these three levels, we introduce fuzziness into the resolution principle and unification and cause fuzzy logical inference to be significant in Fuzzy Prolog when the values of all variables are taken in the closed interval [0, 1].

The automatic training of rule bases that use numerical uncertainty representations : Richard A. Caruana Philips Laboratories, North American Philips Corporation, 345 Scarborough Road, Briarcliff Manor, New York 10510

The use of numerical uncertainty representations allows better modeling of some aspects of human evidential reasoning. It also makes knowledge acquisition and system development, test, and modification more difficult. We propose that where possible, the assignment and/or refinement of rule weights should be performed automatically. We present one approach to performing this training - numerical optimization - and report on the results of some preliminary tests in training rule bases. We also show that truth maintenance can be used to make the training more efficient and ask some epistemological questions raised by training rule weights.

A new renormalisation-group weight function for linear polymers

The author proposes a true centre rule weight function for treating linear polymers with a cell position-space renormalisation group (PSRG). For a cell-to-cell PSRG, the preliminary results for small cells give the critical fugacity per monomer K*=0.3886 and the critical exponent nu =0.7471. Accurate results are also obtained for nu with a cell-to-bond PSRG.

Algorithm 417: Rapid computation of weights of interpolatory quadrature rules

This algorithm is a companion to [1] where the theoretical background is described

The symmetry groups of non-rigid molecules

The concept of molecular symmetry is extended to molecules such as ethane and hydrazine which can pass from one conformation to another. The symmetry group of such a molecule is the set of (i) all feasible permutations of the positions and spins of identical nuclei and (ii) all feasible permutation-inversions, which simultaneously invert the coordinates of all particles in the centre of mass. According to the representations of this group one can classify not only the spin states and states of motions of the nuclei, but even the electronic states of the molecule. Examples are given to illustrate the use of this concept in determining the statistical weights of individual levels and selection rules for electric dipole transitions between them.