Vol. 27 · No. 3

The notion of continuity from above (upper continuity) for lattice-valued possibilistic measures as investigated in cite7 has been proved to be a rather strong condition when imposed as demand on such a measure. Hence, our aim will be to introduce some versions of this upper continuity weakened in the sense that the conditions imposed in cite7 to the whole definition domain of the possibilistic measure in question will be restricted just to certain subdomains. The resulting notion of partial upper convergence and continuity of lattice-valued possibilistic measures will be analyzed in more detail and some results will be introduced and proved.

This paper considers the problem of mining recent frequent itemsets over data streams. As the data grows without limit at a rapid rate, it is hard to track the new changes of frequent itemsets over data streams. We propose an efficient one-pass algorithm in sliding windows over data streams with an error bound guarantee. This algorithm does not need to refer to obsolete transactions when they are removed from the sliding window. It exploits a compact data structure to maintain potentially frequent itemsets so that it can output recent frequent itemsets at any time. Flexible queries for continuous transactions in the sliding window can be answered with an error bound guarantee.

This paper exposes a research of the NP-hard Maximally Balanced Connected Partition problem (MBCP). The proposed solution comprises of a genetic algorithm (GA) that uses: binary representation, fine-grained tournament selection, one-point crossover, simple mutation with frozen genes and caching technique. In cases of unconnected partitions, penalty functions are successfully applied in order to obtain the feasible individuals. The effectiveness of presented approach is demonstrated on the grid graph instances and on random instances with up to 300 vertices and 2 000 edges.

Most learning algorithms assume that all the relevant data are available on a single computer site. In the emerging networked environments learning tasks are encountering situations in which the relevant data exists in a number of geographically distributed databases that are connected by communication networks. These databases cannot be moved to other network sites due to security, size, privacy, or data-ownership considerations. In this paper we show how a k-nearest classifier algorithm can be adapted for distributed data situations. The objective of our algorithms is to achieve the learning objectives for any data distribution encountered across the network by exchanging local summaries among the participating nodes.

The current methodology for designing highly efficient technological systems needs to choose the best combination of the parameters that affect the performance. In this paper we propose a promising optimization algorithm, referred to as the Multilevel Ant Stigmergy Algorithm (MASA), which exploits stigmergy in order to optimize multi-parameter functions. We evaluate the performance of the MASA and Differential Evolution – one of the leading stochastic method for numerical optimization – in terms of their applicability as numerical optimization techniques. The comparison is performed using several widely used benchmark functions with added noise.

In this article a modification of a grammar systems theoretic construction, the so-called network of parallel language processors, is proposed to describe the behaviour of peer-to-peer (P2P) systems. In our model, the language processors form teams, send and receive information through collective and individual filters. The paper deals with the dynamics of string collections. The connection between the growth function of a developmental system and the growth function of networks of parallel multiset string processors with teams of collective and individual filtering is also established.

A method for segmentation of cardiac structures especially for mitral valve in echocardiographic sequences is presented. The method is motivated by the observation that the structures of neighboring frames have consistent locations and shapes that aid in segmentation. To cooperate with the constraining information provided by the neighboring frames, we combine the template matching with the conventional snake model. It means that the model not only is driven by conventional internal and external forces, but also combines an additional constraint, the matching degree to measure the similarity between the neighboring prior shape and the derived contour. Furthermore, in order to automatically or semi-automatically segment the sequent images without manually drawing the initial contours in each image, generalized Hough transformation (GHT) is used to roughly estimate the initial contour by transforming the neighboring prior shape. Based on the experiments on forty sequences, the method is particularly useful in case of the large frame-to-frame displacement of structure such as mitral valve. As a result, the active contour can easily detect the desirable boundaries in ultrasound images and has a high penetrability through the interference of various undesirables, such as the speckle, the tissue-related textures and the artifacts.