Abstract. The main problem of task processing in distributed hard real-time systems is to guarantee execution of all tasks before termination of their deadlines. Scheduling of such tasks is based on their characteristics, which are known a priori. This paper presents a decentralized task scheduling algorithm for distributed hard real-time systems. In this algorithm a new scheme of interactions between one scheduler module and scheduler modules on other nodes is employed. The proposed scheme, defined here as node addressing, is simpler than a focussed addressing scheme or a bidding scheme which are described in the literature. It allows dynamic scheduling of tasks and, moreover, contrary to previously mentioned schemes, makes possible to extend easily a distributed system without interfering with the software of remaining nodes.

Abstract. An important part of a distributed hard real-time system design is the choice of a scheduling algorithm. We evaluate a distributed scheduling algorithm presented in the paper (A Decentralized Task Scheduling Argument for Distributed Hard Real-Time Systems, presented in this issue). The results of the study show that the parameters of the investigated algorithm are comparable with others. Proposed algorithm may provide an effective solution of distributed scheduling in hard real-time environment.

Abstract. The unit-height rectangle packing problem (known also as the on-line interval graph coloring) is investigated. It is proved that the worst case performance ratio of the first-fit strategy is not less than 4.45. This improves the previous bound of 4 established by Witsenhausen, Chrobak, and Œlusarek.

Abstract. The TR'System supervises Computer Teaching Room users' work and keeps track of their results, especially those of didactical importance. A teaching schedule, which is defined for each student, specifies which lessons (educational programs), and under what conditions, should be carried out. The student is given much freedom in choosing lessons to work with, but the menu offered for a subsequent session depends heavily on the results obtained so far. A teacher may update schedule parameters between student's sessions, thus adapting its difficulty to his abilities. The System provides means for adjustment of any educational program to the TR environment, hence there are no restrictions on didactical methods used. Ordinary users need not have any computer experience to work with the TR'System, so its possible application ranges from primary school up to university.

Abstract. The family of directed node- and edge-labelled graphs (EDG-graphs) is a convenient tool for describing two-dimensional patterns. Defining an unambiguous graph representation is a problem which makes difficult the use of graph grammars for syntactic pattern recognition. An algorithm indexing EDG-graphs presented in the paper is based on structural features of patterns represented by these graphs. The family of indexed EDG-graphs is a formalism convenient for further processing in a pattern recognition system.

Abstract. BND - Nagy's method of feature extraction as an algebraic structure is presented in the paper. A description of feature extraction operations treated as a sequence of binary operations on a certain group is introduced. Then, such a description is used for complex feature extraction.

Abstract. In this paper we continue the considerations of (n,p)-forest languages introduced in [2] (Languages and Systems of (n,p)-Forests,) and [3] (Regular Expressions for Languages of (n,p)-Forest) (see: Schedae Informaticae, Issue 4 (1991)). We define a valued matrix and show that it is a matrix representation of a forest. A matrix correspondence for all results from paper [2] and [3] are obtained.

Abstract. An edNLC-graph grammar (introduced in Janssens and Rozenberg) is a strong graph generative formalism. The problem of syntax analysis (parsing) for this graph grammar has been solved. In this paper a complete solution of a grammatical inference problem for edNLC-graph grammars is presented.

Abstract. This paper analyses a certain problem from statistical pattern recognition (classification), namely the application of the gamma distribution in supervised parametric learning. Generally, the a posteriori density function produced by the application of Bayes' rule does not possess a simple functional form, and often can only be evaluated numerically. This paper shows that for some special cases the computation of the a posteriori density function by the application of Bayes' rule is reduced to updating parameters of of the gamma distribution each time a new observation is obtained.

Abstract. This paper analyses a certain problem from statistical pattern recognition (classification), namely the application of the beta distribution in supervised parametric learning. Generally, the a posteriori density function produced by the application of Bayes' rule does not possess a simple functional form, and often can only be evaluated numerically. This paper shows that for some special cases, the computation of the a posteriori density by the application of Bayes' rule is reduced to updating parameters of of the beta distribution each time a new observation is obtained.