Showing papers presented at "Computer Aided Systems Theory in 1992"

A machine learning-based approach to load balancing in computer networks

Abstract: This paper presents an AI-based approach to Load balancing in computer networks; the load redistribution in the network is controlled by a rule-based expert system. Our experiments have shown that changes in the workload structure impose the necessity of automated modifications of the rules in the knowledge base. These modifications are made by means of a machine learning subsystem that is incremental and capable of forgetting pieces of knowledge that have become obsolete. This paper can be understood as a brief report on a real-world application of Michalski's idea of flexible concepts.

A systematic method for exploring contour segment descriptions

Abstract: A new methodology has been developed for the comparative analysis of different contour characterization methods in the context of a structural approach for an artificial vision system. This methodology has been designed for testing the description of contour segments using characterization methods that have been used in applications dealing with isolated forms. The proposed schema is modular and comprises three major steps: a segmentation stage, which is based on a rule-based segmentation machine; a labeling stage, which is dependent on the characterization method used to describe the contour segments and in which the problem of defining classes or typologies of segments is also considered; and a learning and recognition stage, which uses fast tree classifiers and combines their evidence by means of the Dempster-Shafer combination rule. Experimental results are presented using the Fourier-Bessel transform as the contour characterization method.

Robot path planning in the configuration space with automatic obstacle transformation

Abstract: A method of solving the problem of finding the set of movements that allows a robot to go from a starting to an ending position while avoiding obstacles consists of using a path searcher in the configuration space. In order to achieve this, the complete space is divided into two disjoint subspaces. One of the two, the forbidden space, contains all the configurations leading to collisions; the other, the free space, contains all the configurations avoiding obstacles. The problem to be considered now is finding a path in the free subspace to maximize an optimal criterion. The generation of the free space is difficult to solve, and it is usually handled in an approximate way to permit an analytical treatment. Here we present a method of automatically generating the configuration set corresponding to an obstacle. The system is complemented by a path finder based on an algorithm of heuristic searching of solutions in state spaces.

Improving neural net training by mathematical optimization

Abstract: Multilayer perceptrons are popular neural net models that are frequently used for classification tasks. Unfortunately, the standard training methods (backpropagation algorithm and variants thereof) are often numerically inefficient and unstable and need ad hoc tuning. In contrast to these methods, in this paper mathematically profound optimization techniques are applied: a combination of conjugate gradient methods, quasi-Newton methods, and several line search algorithms. This leads to considerable improvements over the backpropagation algorithm on various training data sets (parity problem, phoneme classification, recognition of handwritten characters).

A parallel algorithm for householder's tridiagonalization of a symmetric matrix

Abstract: The tridiagonalization of symmetric matrices appears frequently in the solution of mathematical problems. In this paper we present the parallelization of Householder’s method in hypercube computers with distributed memory and messagepassing communications, using a single program-multiple data mode of parallelism. Results for the NCUBE/10 and for a SUPERNODE system, composed of eight transputers T800 interconnected as a hypercube, are included. We have designed a parallel algorithm getting minimum data redundancy (the matrices are symmetric) and a very high computational load balance among the processing elements. © 1992 Taylor & Francis Group, LLC.

Asynchronous dialogue control in a medical graphic interface

Abstract: Graphic interfaces facilitate the implementation of asynchronous dialogue between human and computer. This paper presents a graphic interface for an application in medicine, based on the declarative definition of graphic templates. The templates are specified from predefined types, and the introduction of data into a template produces a graphic object on the screen. The representational model we have used for the specification of behavior corresponds to finite state automata. Each template has an associated transition table. The control of the interaction process is distributed among the templates. There are an external dialogue between the user and the graphic objects and an internal dialogue between the application programs and the templates. The approach we present is flexible enough to permit the design of an interaction module for real use in a fast and simple way and facilitates iterative design, attaching more importance to the consideration of human factors than to programming. © 1992 Taylor & Francis Group, LLC.

The application of structured spreadsheet modeling in medicine

Abstract: The effective use of computers in medical praxis is limited first by difficulties of medical staff in learning and using conventional design paradigms and second by the variety of medical applications. Only when medical staff can use computers quickly, easily, and in a natural, intuitive fashion will medical computer-supported systems be used enthusiastically and successfully. In the present paper we show that structured spreadsheet modeling, which is a combination of structured modeling and spreadsheet software, can be used effectively and efficiently in the process of designing, testing, and evaluating various classes of medical models.

The importance of estimation and approximation methods in system theory

Abstract: In several fields of cybernetics—for instance, in measurement—the input signals are not known; that means they are to be supposed using the a priori information of the process. Therefore estimation and approximation methods play an important role. In this paper some typical examples will be treated: First, estimation of the errors caused by violations of die sampling theorem (cutoff and aliasing errors) is treated. It is shown that in particular modern sensors with direct digital output often produce aliasing errors because of the direct coupling of the input of the sensor to die signal source. On the other hand, these errors are dependent on the processing after sampling. The next example concerns problems of optimal filtering. Because in practice the power density of a signal is often not known, approximations are used. Furthermore, it is shown that limitations of the degree of correction arise from changing parameters of the system. This sensitivity of the corrected system is greater as the de...

Adaptive method for image segmentation based in local feature

Abstract: This paper presents a method for finding a threshold surface that fits each region of the image and allows separation of objects from background. A surface threshold is determined by interpolating die gray levels at points where there is a high gradient, corresponding with object boundaries. The gradient map and die gray-level distribution give the necessary information for calculating an adaptive threshold level at each point of the image that allows successful segmentation of images in which object-background boundaries are well defined. This method has been tested successfully in different applications of artificial vision oriented to object manipulation by robots and die measurement of wood volume carried by trucks in a cellulose factory.

Casual Bayesian reasoning in medicine

Abstract: We propose a knowledge representation architecture organized in three levels—a causal network (containing the domain knowledge), medical strategies and causal reasoning—and expose the fundamentals of a mathematical model for computing probaility through the application of Bayes' theorem in a causal network.

Strategic knowledge in an expert system for agriculture

Abstract: Most of the expert systems applied in vegetal pathology treat the problem of selecting treatment in a conventional manner, by means of production rules that associate to each pathology the most adequate chemical product. This makes it difficult to generate useful explanations. In order to generate satisfactory explanations the knowledge of the system must be based on the strategies used by human experts. This article introduces our approach for the identification and representation of strategic knowledge in an expert system for plague control in greenhouses. We present an introduction to the application domain and make an analysis of the strategic knowledge implied. We distinguish between the underlying strategy and the practical strategy used by the expert for the solution of the problem. From this we propose a preliminary architecture based on strategic reasoning agents.

Cognitive psychology and dynamical systems theory

Abstract: This paper tries to give a system-theoretical foundation to cognitive psychology, starting from the theoretical and experimental work done in the departments of psychology and mathematics of the Rome University “La Sapienza” by a team formed by M. Olivetti Belardinelli, E. Pessa, V. De Pascalis, C. Del Miglio, and F. S. Marucci. It contains a new mathematical formulation of the organism-environment relationship, a new global brain model, a theory of visual perception based on higher-order neural networks, and reports of experimental observations of phase transitions taking place in the cognitive system.

Image coding through progressive resolution transforms

Abstract: Visual information must be represented digitally to allow its processing by computer. This representation by means of a finite amount of digital data conveys an enormous amount of superfluous information that can be eliminated. We can consider two kinds of superfluous information: statistical redundant and subjective redundant. Discrete transform coding takes blocks of pixels and transforms them into another domain, the transform domain, prior to coding and transmission. An important property is that coefficients need not to be transmitted all in order to obtain good-quality reconstructions. We have applied a special nonor-thogonal transform in order to compress radiographic images.

Theory and algorithms on period assignment of discrete-event dynamic systems

Abstract: For a discrete-event dynamic system described by max algebra as X(£) −X(k − 1)A ⌖ V(k)B, all the eigenvalues λi 1 ≤ i ≤ ω in the right upper triangular standard form of A are called periods. It is proved in this paper that the periods can be assigned arbitrarily in [λI *plus; ∞) by state feedback U(k) = X(k)K if and only if A and B have matching standard forms. Algorithms on period assignment are also given.

FC: a relational query language having the property of functional completeness

Abstract: Functional completeness of a relational language is the ability to express linear recursive queries. We present such a language that also has the property of relational completeness (relational algebra is, in fact, embedded in it). The transitive closure of a binary relation is carried out by means of the projection function and its inverse.

A hybrid architecture for distributed autonomous control agents

Abstract: Business environments, like most other real systems, are distributed in nature. However, the fact that the environment itself is distributed does not justify a distributed control architecture. On the other hand, a centralized architecture is not exactly an ideal form of control. The architecture proposed herein is a hybrid control architecture that maps the decision-making environment of many situations ranging from factory floors to traffic control. The architecture assumes the presence of self-reliant (autonomous) decision-making agents, which are required to respond in a certain amount of time. Cooperation between these agents is assumed, and therefore each agent is expected to perform as best as it can in order to complete a well-defined task. No central coordination or control agent is assumed. Thus an asynchronous behavior is assumed.