Traditional information retrieval evaluation relies on both precision and recall. However, modern search environments such as the Web, in which recall is either unimportant or immeasurable, require precision-oriented evaluation. In particular, finding one highly relevant document is very important for practical tasks such as known-item search and suspected-item search. This paper compares the properties of five evaluation metrics that are applicable to the task of finding one highly relevant document in terms of the underlying assumptions, how the system rankings produced resemble each other, and discriminative power. We employ two existing methods for comparing the discriminative power of these metrics: The Swap Method proposed by Voorhees and Buckley at ACM SIGIR 2002, and the Bootstrap Sensitivity Method proposed by Sakai at SIGIR 2006. We use four data sets from NTCIR to show that, while P(⁺)-measure, O-measure and NWRR (Normalised Weighted Reciprocal Rank)are reasonably highly correlated to one another, P(⁺)-measure and O-measure are more discriminative than NWRR, which in turn is more discriminative than Reciprocal Rank. We therefore conclude that P(⁺)-measure and O-measure, each modelling a different user behaviour, are the most useful evaluation metrics for the task of finding one highly relevant document.

View full abstract

We model the relationships between the message formats of a business system and their semantics in a machine-processable knowledge base. We describe a message-mapping technique that extracts the relationships between the message formats of several systems semiautomatically by using the class characteristics of the semantics and stores these relationships as past system design knowledge. In addition, we propose process-mapping, which is a technique that discovers suitable software components for system cooperation. We evaluate these methods using the interface specifications of actual service management systems and show that the frequency of interface adjustment can be reduced.

View full abstract

Searching in a spatial database for 3D objects that are similar to a given object is an important task that arises in a number of database applications, for example, in medicine and CAD fields. Most of the existing similarity searching methods are based on global features of 3D objects. Developing a feature set or a feature vector of 3D object using their partial features is a challenging. In this paper, we propose a novel segment weight vector for matching 3D objects rapidly. We also describe a partial and geometrical similarity based solution to the problem of searching for similar 3D objects. As the first step, we split each 3D object into parts according to its topology. Next, we introduce a new method to extract the thickness feature of each part of every 3D object to generate its feature vector and a novel searching algorithm using the new feature vector. Finally, we present a novel solution for improving the accuracy of the similarity queries. We also present a performance evaluation of our stratagem. The experiment result and discussion indicate that the proposed approach offers a significant performance improvement over the existing approach. Since the proposed method is based on partial features, it is particularly suited to searching objects having distinct part structures and is invariant to part architecture.

View full abstract

In this paper, an architecture of software environment to offload user-defined software modules to Maestro2 cluster network, named Maestro dynamic offloading mechanism (MDO), is described. Maestro2 is a high-performance network for clusters. The network interface and the switch of Maestro2 have a general-purpose processor tightly coupled with a dedicated communication hardware. MDO enables the users to offload software modules to both the network interface and the switch. MDO includes a wrapper library with which offload modules can be executed on a host machine without rewriting the program. The overhead and the effectiveness of MDO are evaluated by offloading collective communications.

View full abstract

The matching of a bipartite graph is a structure that can be seen in various assignment problems and has long been studied. The semi-matching is an extension of the matching for a bipartite graph G =(U ∪ V, E). It is defined as a set of edges, M ⊆ E, such that each vertex in U is an endpoint of exactly one edge in M. The load-balancing problem is the problem of finding a semi-matching such that the degrees of each vertex in V are balanced. This problem is studied in the context of the task scheduling to find a “balanced” assignment of tasks for machines, and an O(¦E¦¦U¦) time algorithm is proposed. On the other hand, in some practical problems, only balanced assignments are not sufficient, e.g., the assignment of wireless stations (users)to access points (APs) in wireless networks. In wireless networks, the quality of the transmission depends on the distance between a user and its AP; shorter distances are more desirable. In this paper, We formulate the min-weight load-balancing problem of finding a balanced semi-matching that minimizes the total weight for weighted bipartite graphs. We then give an optimal condition of weighted semi-matchings and propose an O(¦E¦¦U¦¦V¦) time algorithm.

View full abstract

This work proposes a method to control the dominance area of solutions in order to induce appropriate ranking of solutions for the problem at hand, enhance selection, and improve the performance of MOEAs on combinatorial optimization problems. The proposed method can control the degree of expansion or contraction of the dominance area of solutions using a user-defined parameter S. Modifying the dominance area of solutions changes their dominance relation inducing a ranking of solutions that is different to conventional dominance. In this work we use 0/1 multiobjective knapsack problems to analyze the effects on solutions ranking caused by contracting and expanding the dominance area of solutions and its impact on the search performance of a multi-objective optimizer when the number of objectives, the size of the search space, and the feasibility of the problems vary. We show that either convergence or diversity can be emphasized by contracting or expanding the dominance area. Also, we show that the optimal value of the area of dominance depends strongly on all factors analyzed here: number of objectives, size of the search space, and feasibility of the problems.

View full abstract

Immersive virtual reality (VR) has long been considered an excellent environment in which to manipulate 3D virtual objects. However currently used immersive VR user interfaces have limitations. For example, while direct manipulation by hand is easy to understand and to use for approximate positioning, direct manipulation by hand is not suitable for making fine adjustments to virtual objects in an immersive environment because it is difficult to hold an unsupported hand in midair and then to release an object at a fixed point. We therefore propose a method that combines direct 3D manipulation by hand with a virtual 3D gearbox widget that we recently designed. Using this method, hand manipulation is used first to move virtual objects and place them in an approximate position, and then the widget is used to move them into a precise position. The experimental evaluation showed that this combination of direct manipulation by hand and the proposed gearbox is the best of five tested methods in terms of completion ratio of task and subjective preference.

View full abstract

In many cases of biological observations such as cell array, DNA micro-array or tissue microscopy, primary data are obtained as photographs. Specialized processing methods are needed for each kind of photographs because they have very wide variety, and often needed automated systems for modern high-throughput observations. We developed a fully-automated image processing system for cell array, high-throughput time series observation system for living cells, to evaluate gene expression levels and phenotype changes in time of each cell.

View full abstract

In many of the chemical reactions in living cells, enzymes act as catalysts in the conversion of certain compounds (substrates) into other compounds (products). Metabolic pathways are formed as the products of these reactions are used as the substrates of other reactions. Comparative analyses of the metabolic pathways among species provide important information on both evolution and potential pharmacological targets. Here, we propose a method to align the metabolic pathways based on similarities between chemical structures. To measure the degree of chemical similarity, we formalized a scoring system using the MACCS keys and the Tanimoto/Jaccard coefficients. To determine the effectiveness of our method, it was applied to analyses of metabolic pathways in Escherichia coli. The results revealed compound similarities between fructose and mannose biosynthesis and galactose biosynthesis pathways.

View full abstract

Tree Adjoining Grammar (TAG) is a useful grammatical tool to model RNA secondary structures containing pseudoknots, but its time complexity for parsing is not small enough for the practical use. Recently, Weinberg and Ruzzo proposed a method of approximating stochastic context free grammar by stochastic regular grammar and applied it to faster genome annotation of non-coding RNA families. This paper proposes a method for extending their idea to stochastic approximation of TAGs by regular grammars. We will also report some preliminary experimental results on how well we can filter out non candidate parts of genome sequences by using obtained approximate regular grammars.

View full abstract

We created a Free Energy Landscape Analysis System based on a parallelized molecular dynamics (MD) simulation adapted for the IBM Blue Gene/L supercomputer. We begin with an outline of our Free Energy Landscape Analysis system. Next we discuss how Parallel MD was tuned for Blue Gene/L. We then show the results for some test targets run on Blue Gene/L, including their efficiency. Finally, we mention some future directions for extension of this project.

View full abstract

This paper investigates the interaction ability when introducing pressure into current basic interaction techniques by developing two novel techniques. A Zoom-based technique with pressure (hereafter referred to as ZWPS) is proposed to improve pixel-target selection. In this technique the pressure is used as a switch mode to couple a standard Point Cursor and a zoomable technique together. Pressure Scroll is also presented with a view to advancing scrolling performances by employing arc or line strokes to scroll documents. In this technique pressure is used as an additional control factor to widen the adjustable range of the scrolling velocity. We conducted two experiments to examine the effectiveness of ZWPS and Pressure Scroll. The experimental results indicate that they both bring significant benefits to the users.

View full abstract

The QoS provided by current best effort Internet is not good enough for real-time multimedia applications that are categorized as premium traffic. It is believed that QoS guarantees could be better provided by the connection oriented networks. Multi Protocol Label Switching (MPLS) is one such technology and these connection oriented networks are inherently more prone to network failures. Re-routing is a solution to cope with them. However, the re-routing always causes packet losses and results in service outage. Therefore, the QoS of the real-time premium traffic is highly degraded. The seamless failure recovery proposed in this paper can be used for real-time premium traffic that needs a guaranteed QoS. It applies an FEC technique to the conventional re-routing based path protection and seamlessly recovers the packet losses due to re-routing by way of an FEC recovery technique. The numerical and simulation results show that the proposed method can provide network architecture without service outage for real-time premium traffic while minimizing the service costs such as redundant traffic and additional buffer at edge routers.

View full abstract

This paper outlines some of the benefits of providing remote users with consistent spatial referencing across sites when collaborating on physical tasks. Two video-mediated technologies are introduced: room-sized sharing that enables remote users to see similar things to what they would actually see if in the same room and a snapshot function that enables users to gesture at remote objects. We examine the impact of these technologies by comparing remote collaboration on physical tasks in a regular video conferencing system with a handy camera versus a room duplication system versus a room duplication system with a snapshot function. Results indicate that room-sized sharing facilitates remote collaborators' sense of co-presence and supports remote gesturing, which is closely aligned to normal co-present gesturing. Although such benefits did not contribute directly to the overall decrease of task performance, room-sized sharing and the snapshot function helped remote collaborators construct appropriate messages, efficiently establish joint focus, and monitor each others' comprehension when conducting complicated physical tasks.

View full abstract

A recent trend in computing has been to leverage dormant PC resources. To achieve this, background applications such as peer-to-peer applications and PC Grid run on ordinary users' PCs, sharing their computing resources. If not properly managed, the background applications obtrude on the PC user's active jobs. In particular, the contention over disk bandwidth severely degrades performance. In this paper, we present DiscNice, a novel scheme for disk bandwidth management that can host background applications unobtrusively. Its novelty lies in the fact that it throttles disk I/O completely at the user-level. The user-level approach is attractive for heterogeneous environments such as differently configured PCs over the world; portability is enhanced and deployment is easier in comparison with kernel-level approaches. Experimental results suggest that our prototype DiscNice running on Linux 2.4.27 incurs 12% or less overhead, and gracefully ensures the unobtrusiveness of background applications.

View full abstract