Catálogo de publicaciones - libros

Over the last several years there has been an explosion in the number of computational methods for the detection of transcription factor binding sites in DNA sequences. Although there has been some success in this field, the existing tools are still neither sensitive nor specific enough, usually suffering from the detection of a large number of false positive signals. Given the properties of genomic sequences this is not unexpected, but one can still find interesting features worthy of further computational and laboratory bench study. We present an efficient algorithm developed to find all significant variable motifs in given sequences. In our view, it is important that we generate complete data, upon which separate selection criteria can be applied depending on the nature of the sites one wants to locate, and their biological properties. We discuss our algorithm and our supplementary software, and conclude with an illustration of their application on two eukaryotic data sets.

Shopping in the Internet does not proceed smoothly as customers expect currently. In particular customers have many obstacles in finding an optimal combination of products and shopping malls to minimize the total purchasing cost. To solve such problems, this paper proposes a new framework of an intelligent shopping agent based on the Semantic Web and the Integer Programming technologies. Starting from the search of products, it will show how to build an intelligent agent by using concepts of the Semantic Web and how to connect information to formulating and solving an optimization problem to achieve a customer’s goal.

This paper describes the implementation of an indexing mechanism on a Rete-based reasoner working with ontological data in order to optimize memory consumption of the reasoner. This newly introduced indexing mechanism is known as the Pyramid Technique [1]. Our work organizes three dimensional ontological data in a way that works efficiently with this indexing mechanism and it constructs a subset of the querying scheme of the Pyramid Technique that supports querying ontological data. This work also implements an optimization on the Pyramid Technique. Finally, it discusses the performance analysis of the reasoner in terms of time and memory consumptions.

Calculation of the required engine power and displacement takes an important place in the initial design of motorboats. Recently, several calculation methods with fewer parameters and with a possible gain of time compared to classical methods have been proposed. This study introduces a novel calculation method based on neural networks. The method requires less data input and hence is more easily applicable than classical methods. In this study several different neural network methods have been conducted on data sets which have principal parameters of motorboats and the respective performances have been presented. From the results obtained, displacement and engine power prediction for motor boats can be used at a suitable level for ship building industry.

In this paper, we introduce a dynamic role assignment mechanism for a team of cooperative virtual agents working in interactive computer games. This role assignment mechanism is a new one different from both existing static and dynamic mechanisms, in which decisions regarding role assignment are made all at once either in the design phase or in the execution phase. According to our mechanism, all situation-dependent role sets are predefined in the design phase. Detail decisions regarding which member agent has to take what specific role, however, are made in the execution phase.  This mechanism can minimize the negotiation effort for role assignment in the execution phase. Therefore, this mechanism is quite effective in real-time multi-agent environments like interactive computer games. Through experiments, we show the superiority of our dynamic role assignment mechanism.

This paper presents a statistical lexical ambiguity resolution method in direct transfer machine translation models in which the target language is Turkish. Since direct transfer MT models do not have full syntactic information, most of the lexical ambiguity resolution methods are not very helpful. Our disambiguation model is based on statistical language models. We have investigated the performances of some statistical language model types and parameters in lexical ambiguity resolution for our direct transfer MT system.

An important problem in face recognition is the design of the feature space which represents the human face. Various feature sets have been and are continually being proposed for this purpose. However, there exists no feature set which gives a superior and consistent recognition performance on various face databases. Concatenating the popular features together and forming a high dimensional feature space introduces the curse of dimensionality problem. For this reason, dimensionality reduction techniques such as Principal Component Analysis is utilized on the feature space. In this study, first, some of the popular feature sets used in face recognition literature are evaluated over three popular face databases, namely ORL [1], UMIST [2], and Yale [3]. Then, high dimensional feature space obtained by concatenating all the features is reduced to a lower dimensional space by using the Minimal Redundancy Maximal Relevance [4] feature selection method in order to design a generic and successful feature set. The results indicate that mRMR selects a small number of features which are satisfactory and consistent in terms of recognition performance, provided that the face database is statistically stable with sufficient amount of data.

This paper discusses the effects of the prediction of return addresses in high-performance processors designed with trace caches. We show that a traditional return address stack used in such a processor predicts return addresses poorly if a trace cache line contains a function call and a return. This situation can often be observed for processors demanding aggressive instruction fetch bandwidth. Thus, we propose two potential schemes to improve the prediction accuracy of return addresses. We demonstrate that the proposed schemes increase the return address prediction rates reasonably using minimal hardware support. We also analyze the effects of various trace cache configurations on the return address prediction accuracy such as trace cache set associativity, cache size and line size. Our experimental results show that the average return address prediction accuracy across several benchmarks can be up to 11% better than a traditional return address stack in a high-performance processor with a trace cache.

Correct branch prediction is an essential task in modern microarchitectures. In this paper, to additionally increase the prediction accuracy, recovery logics for speculative update branch history are presented. In local or global branch predictors, maintaining speculative update history provides substantial prediction accuracy. However, speculative update history requires a suitable recovery mechanism. This paper proposes recovery logics for speculative update branch history, for both global- and local-history. The proposed solutions provide higher prediction accuracy and guarantee the correctness of program, and they can be efficiently implemented with low hardware costs.

One of the main difficuties to map an embedded application onto a multiprocessor architecture is that there are multiple ways of this mapping due to several constraints. In this paper, we present an Integer Linear Programming based framework that maps a given application (represented as a task graph) onto a Heterogeneous Chip Multiprocessor architecture. Our framework can be used with several objective functions such as energy, performance, and fallibility (opposite of reliability). We use Dynamic Voltage Scaling (DVS) for reducing energy consumption while we employ task duplication to minimize fallibility. Our experimental results show that over 50% improvements on energy consumption are possible by using DVS, and the fully task duplicated schedules can be achieved under tight performance and energy bounds.