Volume 15, 1996, No. 5


Cognitivism and Situated Action: Two Views on Intelligent Agency

N. Kushmerick

Abstract. Classical approaches to artificial intelligence have in recent years been challenged by several alternative approaches. This review article focuses on one such alternative, which has come to be called 'Situated Action' (SA). I analyze  recent research allied with this alternative approach, and identify five major themes that distinguish SA from more traditional research. I conclude that though much additional work is needed in order to fully elaborate the SA conception of intelligence and agenda, Situated Action does indeed suggest a number of novel research directions, as well as provide important accounts of some aspects of agency that are problematic from the classical perspective.

 

Grammar Systems with WAvE-like Communication

E. Csuhaj-Varjú, J. Kelemen, Gh. Paun

Abstract. We propose a series of variants of parallel communicating grammar systems (PC grammar systems) which are able to simulate data flow in WAVE/like architectures of parallel processors. In contrast to the PC grammar systems investigated so far, where the communication is done by request, here the communication is done by command,  the target processors being selected according to the pattern of the string to be communicated. Motivated by suggestions from the WAVE paradigm, the Boltzmann machines, the Connection Machine and other parallel devices, we informally consider a series of variants of PC grammar systems. Then we formally define and investigate one of them, in order to demonstrate its computational power and mathematical fruitfulness.  Further investigations remain to be conducted in order to recognize which of these variants are more useful from theoretical and practical points of view than the others.

 

An Object-Oriented Concurrent Language for Extensibility and Reuse of Synchronization Components

L. Thomas

Abstract. Concurrency in object oriented languages seems natural. However well known problems arise when we try to define such concurrent object oriented languages. These problems deal with interferences between expressing true concurrency inside objects and expressing re-use.  This article presents a concurrent object model and illustrates through various examples how it solves these interferences.

 

Note on Problems which are Hard for some Weakly Connected Parallel Architectures

I. Trenčanský

Abstract. In this paper the lower bound technique, based on information content, for special models of VLSI circuits defined by some topological restrictions is investigated. The assertion bounding possibilities of speeding up a VLSI  computation by increasing the number of processors for circuits with f/separators is presented. Further possibilities of applying these results to obtain stronger lower bounds or proofs of noneffectivity of speeding up computation for some classes of problems and separators are shown.

 

An Optimal Algorithm for Gaussian Elimination of Band Matrices on a MIMD System

E.I. Milovanovič, I.Ž. Milovanovič, M.K. Stojčev, M.D. Mihajlovič

Abstract. This paper is concerned with determining an optimal number of processors in MIMD system, for LU decomposition of band matrix with a + b - 1 diagonals using Gaussian method of elimination. The obtained result represents the solution of general problem since band matrices with arbitrary number of diagonals were considered. Task scheduling algorithm is given and proved that it is optimal. We introduce orthogonal multiprocessor system with two-dimensional memory organization with r < 16 32/bit processors and r(r-1)/2 shared memory modules. The unique feature of the proposed system lies in this conflict/free access to shared memory modules.

 

A Generalized Fuzzy C-means Algorithm with Applications to Contrast Modification and Binarization of Images

J. Richardt, J. Nicklish-Franken, R. Klette

Abstract. The fuzzy c-means algorithm (FCM) can be applied to several problems in image analysis, ranging from image segmentation [15, 16] to the detection of pictorial patterns [2,3,4,9]. In this paper it is shown that the problems of image binarization and of segmentation of gray  value histograms are closely related to the basic concepts of the FCM. The binarization can be performed by means of "smooth" contrast modifications at several degrees of sharpness. This is due to the fuzzy thresholding technique supplied by the FCM approach. This paper connects fuzzy thresholding with the known sigmoid functions of neural nets, which serve for the same purpose of fuzzy thresholding.  From there a connection arises between the FCM approach and the basic formulas of simulated annealing [12].


Go To Contents