Catálogo de publicaciones - libros

In this paper, a sampling method based on the Metropolis algorithm is proposed. It is able to draw samples that have the same distribution as the underlying probability distribution. It is a simple, efficient, and powerful method suitable for all distributions. We have performed experiments to examine the qualities of the samples by comparing their statistical properties with the underlying population. The experimental results show that the samples selected by our method are bona fide representative.

In a recent editorial of the Bioinformatics journal, Dr. Pavel Pevzner, a pioneering researcher in the field of bioinformatics, made the following statement [1]:

Pattern extraction from music strings is an important problem. The patterns extracted from music strings can be used as features for music retrieval or analysis. Previous works on music pattern extraction only focus on exact repeating patterns. However, music segments with minor differences may sound similar. The concept of the prototypical melody has therefore been proposed to represent these similar music segments. In musicology, the number of music segments that are similar to a prototypical melody implies the importance degree of the prototypical melody to the music work. In this paper, a novel approach is developed to extract all the prototypical melodies in a music work. Our approach considers each music segment as a candidate for the prototypical melody and uses the edit distance to determine the set of music segments that are similar to this candidate. A lower bounding mechanism, which estimates the number of similar music segments for each candidate and prunes the impossible candidates is designed to speed up the process. Experiments are performed on a real data set and the results show a significant improvement of our approach over the existing approaches in the average response time.

Automatic music genre classification has long been an important problem. However, there is a paucity of literature that addresses the issue, and in addition, reported accuracy is fairly low. In this paper, we present empirical study of a novel music descriptor generation method for efficient content based music genre classification. Analysis and empirical evidence demonstrate that our approach outperforms state-of-the-art approaches in the areas including accuracy of genre classification with various machine learning algorithms, efficiency on training process. Furthermore, its effectiveness is robust against various kinds of audio alternation.

Content-based music information retrieval uses features extracted from music to answer queries. For melodic queries, the two main features are the pitch and duration of notes. The note pitch feature has been well researched whereas duration has not been fully explored. In this paper, we discuss how the note duration feature can be used to alter music retrieval effectiveness. Notes are represented by strings called standardisations. A standardisation is designed for approximate string matching and may not capture melodic information precisely. To represent pitches, we use a string of pitch differences. Our duration standardisation uses a string of five symbols representing the relative durations of adjacent notes. For both features, the Smith-Waterman alignment is used for matching. We demonstrate combining the similarity in both features using a vector model. Results of our experiments in retrieval effectiveness show that note duration similarity by itself is not useful for effective music retrieval. Combining pitch and duration similarity using the vector model does not improve retrieval effectiveness over the use of pitch on its own.

Publish/subscribe system captures the dynamic aspect of the specified information by notifying users of interesting events as soon as possible. Fast response time is important for event filtering which requires multiple step processing and is also one of important factors to provide good service for subscribers.

Generally the event arrival rate is time varying and unpredictable. It is very possible that no event arrives in one unit time and multiple events arrive in another unit time. When multiple events with different workloads arrive at the same time, the average response time of multiple-event filtering depends on the sequence of event by event filtering.

As far as we know, significant research efforts have been dedicated to the techniques of single event filtering, they can not efficiently filter multiple events in fast response time. In this paper, we first propose a multiple-event filtering algorithm based on R-tree. By calculating relative workload of each event, event by event filtering can be executed with short-job first policy so as to improve average response time of multiple-event filtering. Furthermore, a self-adaptive model is proposed to filter multiple events in dynamically changing environment.

The multiple-event filtering algorithm and the self-adaptive model are evaluated in a simulated environment. The results show that the average response time can be improved maximum up to nearly 50%. With the self-adaptive model, multiple events can be filtered with average response time always same as or close to the possible best time in the dynamically changing environment.

Filtering is a popular approach to reduce information traffic in subscription-based systems, especially if most subscribers are located on mobile devices. The filters are placed between the subscribers and the subscription server. With increasing number of filters from subscribers, filtering may become a bottleneck and challenge the scalability of such systems. In this paper, we propose to use filter indexing to solve this problem. We propose and study four filter-indexing schemes: Ad-Hoc Indexing Scheme (AIS), Group Indexing Scheme (GIS), Group-Sort Indexing Scheme (GSIS) and B’ Tree Indexing Scheme (BTIS). We evaluate the performance of these four indexing schemes with respect to scalability and other factors. Among the proposed schemes, we find that GSIS is the most efficient indexing scheme for searching and BTIS has the best performance for updating and inserting filters.

Recently, there has been increasing interest in research on push-based information systems that deliver data by broadcast in both wired and wireless environments. This paper proposes new caching strategies to reduce the response time of data access by assuming an environment where clients consecutively issue access requests for multiple data items with think time. The proposed strategies take into account each client’s access characteristics, such as correlation among data items and think-time between a data access and the next access request, and reduce the average response time by caching data items with long expected response time. Moreover, we evaluate the effectiveness of the proposed strategies by simulation experiments.

In the past decade, researchers have combined deductive and object-oriented features to produce systems that are powerful and have excellent modeling capabilities. More recently, an XML query language XTree was proposed. Queries written in XTree are more compact, more convenient to write and easier to understand than queries written in XPath. In this paper, we introduce a novel XML query language XDO2 that extends , with features such as and , and features such as and . Our XDO2 language is more , and to use than current query languages for XML such as XQuery and XPath because it is based on XTree, supports (recursive) deductive rules and the not-predicate. An XDO2 database example is given to motivate the usefulness of the language. The formal treatment of language syntax and semantics are presented in the appendices.

Schema descriptions of XML documents become more and more complicated and schema documents become longer and longer as the structure of XML documents becomes more and more complex. This is mainly because they cannot take full use of object-oriented modeling abilities. In this paper, we extend XML as follows to solve this problem. (1) We extend DTD’s type system to provide richer built-in types. Moreover, a user-defined type can be declared using the mechanism in which an existing type is used as the base type and the set of values represented by the derived type is the subset of values represented by the base type. (2) We extend DTD so that element can be global as well as local. (3) We extend DTD with element hierarchy with nonmonotonic inheritance to support super-element sub-element relationship, overriding of elements or attributes inherited from super-elements, blocking of the inheritance of elements or attributes from super-elements, and conflict handling. (4) We extend XML with polymorphism, which is a fundamental feature in object-oriented data models, to support polymorphic elements, typing of references and polymorphic references. Although we extend DTD to support some key object-oriented features, there is not any syntax change of XML documents to fit for our .