Two major challenges in inferring the sparse topological architecture of Gene Regulatory Networks using computational methods are 1) the low accuracy of predicting connections between genes and 2) the excessive comput...
详细信息
ISBN:
(纸本)9781605581668
Two major challenges in inferring the sparse topological architecture of Gene Regulatory Networks using computational methods are 1) the low accuracy of predicting connections between genes and 2) the excessive computational cost. In order to address these challenges, we have exploited some biological features of yeast cell cycle. One such feature is that, a high proportion of Cell Cycle Regulated genes are periodically expressed;that is genes are maximally expressed to affect and control the regulation of other genes and on completing certain tasks;they are repressed by some other regulator genes. Thus the whole cell cycle progresses systematically through the successive activation and inactivation of CCR genes. To use this feature, we have calculated the peak time of individual genes which falls into one/more phases of the cell cycle. Therefore, genes that peak in the interval of the same phase of the cell cycle have been grouped together. Finally, we have applied the Dynamic Bayesian Network (DBN) algorithm within distinct phases of genes. As a consequence, both the accuracy and the computational cost of our learning algorithm have been improved in comparison with the existing DBN algorithms. Copyright 2009 acm.
A variety of real-world situations are beneficial for wearable computing since it provides information services while users are doing other jobs. Therefore, a simple and hands-free input interfaces have suitable for t...
详细信息
ISBN:
(纸本)9781605581668
A variety of real-world situations are beneficial for wearable computing since it provides information services while users are doing other jobs. Therefore, a simple and hands-free input interfaces have suitable for these computer operations. However, such interface has not been achieved with conventional input devices such as mice or track balls. Although gesture or eye-gaze input techniques have also been developed for wearable computing, they also suffer from problems i.e., a slow pointing speed, difficulty in carrying devices, and complexity in parallel use when doing tasks. We propose a new method of pointing using input of simple gestures with two accelerometers. By dividing the specifications of two coordinates into a combination of two independent motions, we accomplish accurate and intuitive pointing. A user attaches two small accelerometers to both his/her hands or both elbows. The pointing is done by using the intersection of two straight lines, and the movement of the lines is synchronized with that of the accelerometers. In addition, we also propose a method of changing the position of objects being pointed at that is new approach. The results we obtained from our evaluation experiments confirmed that our method was effective. Copyright 2009 acm.
Spectrum-based fault localization is a statistical technique that aims at helping software developers to find faults quickly by analyzing abstractions of program traces to create a ranking of most probable faulty comp...
详细信息
ISBN:
(纸本)9781605581668
Spectrum-based fault localization is a statistical technique that aims at helping software developers to find faults quickly by analyzing abstractions of program traces to create a ranking of most probable faulty components (e.g., program statements). Although spectrum-based fault localization has been shown to be effective, its diagnostic accuracy is inherently limited, since the semantics of components are not considered. In particular, components that exhibit identical execution patterns cannot be distinguished. To enhance its diagnostic quality, in this paper, we combine spectrum-based fault localization with a model-based debugging approach based on abstract interpretation within a framework coined Deputo. The model-based approach is used to refine the ranking obtained from the spectrum-based method by filtering out those components that do not explain the observed failures when the program's semantics is considered. We show that this combined approach outperforms the individual approaches and other state-of-the-art automated debugging techniques. Copyright 2009 acm.
Pervasive computing devices (e.g., sensor networks, localization devices, cameras, etc.) are increasingly present in every aspect of our lives. These devices are able to generate enormous amounts of data, from which k...
详细信息
ISBN:
(纸本)9781605581668
Pervasive computing devices (e.g., sensor networks, localization devices, cameras, etc.) are increasingly present in every aspect of our lives. These devices are able to generate enormous amounts of data, from which knowledge about situations and facts occurring in the world can be inferred;inference can also be done by combining data items and generating new (higher-level) ones. Such data and knowledge is of extreme importance for to context-aware and mobile services. However, we are left with the problem that the possibly huge amount of data and knowledge generated can be very hard to be analyzed and made usable in real-time. The core of the problem in today's pervasive environments lies between the ability to extract meaningful (useful) knowledge from the data while making sure the total amount of data does not become overwhelming to the system. This paper focus on this trade-off using (without loss of generality) the W4 model for contextual data as a case study. Starting from the basic mechanism by which the W4 model autonomously generate new knowledge, the paper shows how this can generate knowledge overflow, and propose a method to select - -in a self-organizing way - -what kinds of knowledge should be generated based on their importance;hence preventing knowledge overflow. Experimental results are reported to support our arguments and proposals. Copyright 2009 acm.
With the development of computer systems, function inlining schemes were used to reduce execution time while increasing codes. In embedded systems such as wireless sensor nodes, there are extreme limitations on memory...
详细信息
ISBN:
(纸本)9781605581668
With the development of computer systems, function inlining schemes were used to reduce execution time while increasing codes. In embedded systems such as wireless sensor nodes, there are extreme limitations on memory space and battery power. This is the reason why function inlining is useful for maximizing memory utilization while minimizing energy consumption of embedded systems. In the previous works, basic inlining schemes were proposed, which were adapted to systems with code memory constraints. However, they were too coarse-grained, and did not evaluate the impact of function inlining in terms of both energy consumption and code memory utilization in actual systems. In this paper, we propose a fine-grained function inlining scheme. We also present the impact of function inlining schemes on resource-constrained embedded systems, in terms of energy consumption and code memory overhead. Based on experimental results, we demonstrate that fine-grained function inlining can improve the energy efficiency of embedded systems while maximizing code memory utilization. Copyright 2009 acm.
Good system performance depends on the correct setting of its configuration parameters. It is observed that such optimal configuration relies on the incoming workload of the system. In this paper, we utilize the Marko...
详细信息
ISBN:
(纸本)9781605581668
Good system performance depends on the correct setting of its configuration parameters. It is observed that such optimal configuration relies on the incoming workload of the system. In this paper, we utilize the Markov decision process (MDP) theory and present a reinforcement learning strategy to discover the complex relationship between the system workload and the corresponding optimal configuration. Considering the limitations of current reinforcement learning algorithms used in system management, we present a different learning architecture to facilitate the configuration tuning task which includes two units: the actor and critic. While the actor realizes a stochastic policy that maps the system state to the corresponding configuration setting, the critic uses a value function to provide the reinforcement feedback to the actor. Both the actor and critic are implemented by multiple layer neural networks, and the error back-propagation algorithm is used to adjust the network weights based on the temporal difference error produced in the learning. Experimental results demonstrate that the proposed learning process can identify the correct configuration tuning rule which in turn improves the system performance significantly. Copyright 2009 acm.
In this paper, we present a novel full cube computation and representation approach, named MCG. In a data cube, each cuboid can be viewed as a set of sub-graphs. In general, redundant sub-graphs are quite common in a ...
详细信息
ISBN:
(纸本)9781605581668
In this paper, we present a novel full cube computation and representation approach, named MCG. In a data cube, each cuboid can be viewed as a set of sub-graphs. In general, redundant sub-graphs are quite common in a data cube, but their elimination is a hard problem as some previous cube approaches demonstrate. The MCG approach differentiates significantly from previous approaches since it efficiently eliminates all common sub-graphs from the entire cube, based on an exact sub-graph matching solution. We propose a matching function to guarantee one-to-one mapping between sub-graphs. The function is computed incrementally, in a top-down fashion, and its computation uses a minimal amount of information to generate unique results, regardless of whether we are using distributive, algebraic or holistic measures. MCG performance analysis demonstrates a similar runtime when compared to Star approach and very low memory consumption (94 - 98% reduction) when compared to a full cube representation. Copyright 2009 acm.
This paper proposes a new K-View algorithm for texture image classification using rotation-invariant features. These features are statistically derived from characteristic view sets for each texture. Unlike the existi...
详细信息
ISBN:
(纸本)9781605581668
This paper proposes a new K-View algorithm for texture image classification using rotation-invariant features. These features are statistically derived from characteristic view sets for each texture. Unlike the existing K-View algorithm, all the views used are transformed into rotation-invariant features and the K views are selected randomly. In contrast, the existing K-View algorithm uses the K-means algorithm for choosing the K views. In this new algorithm the decision of determining a pixel to which texture class it belong to, is made by considering all the views which consist of the pixel being classified. In order to preserve the primitive information of a texture class as much as possible, the proposed algorithm randomly selects k views of the view set from each sample sub-image as the characteristic view set. Experimental results show that the proposed algorithm is more robust and accurate compared with the results of the existing K-View algorithm. Copyright 2009 acm.
Vehicular ad hoc networks (VANETs) are envisaged to become a flexible platform for monitoring road traffic, which will gradually replace more cumbersome fixed sensor deployments. The efficacy of vehicle-assisted traff...
详细信息
ISBN:
(纸本)9781605581668
Vehicular ad hoc networks (VANETs) are envisaged to become a flexible platform for monitoring road traffic, which will gradually replace more cumbersome fixed sensor deployments. The efficacy of vehicle-assisted traffic monitoring systems depends on the freshness of traffic data that they can deliver to users, and the bandwidth used to do so. Clearly, high data freshness will allow users to estimate trip times accurately, and to select the fastest route to a destination. Low bandwidth utilization will allow the traffic monitoring application to coexist symbiotically with a wide variety of vehicle-based applications, ranging from road safety to advertising and entertainment. In this paper, we investigate the problem of minimizing the bandwidth utilization of a vehicle-assisted traffic monitoring system, whilst adhering to user-defined requirements for data freshness. The novelty of our approach is that we jointly optimize two intertwined aspects of traffic monitoring: data acquisition and data forwarding. We investigate how their combined operation trades data freshness for bandwidth utilization, and we propose a novel mechanism that fine-tunes their parameters to optimize the overall system performance. Our mechanism is evaluated using realistic vehicular traces on a real city map. Copyright 2009 acm.
This paper proposes a novel method of achieving fast networking in hosted virtual machine (VM) environments. This method, called socket-outsourcing, replaces the socket layer in a guest operating system (OS) with the ...
详细信息
ISBN:
(纸本)9781605581668
This paper proposes a novel method of achieving fast networking in hosted virtual machine (VM) environments. This method, called socket-outsourcing, replaces the socket layer in a guest operating system (OS) with the socket layer of the host OS. Socket-outsourcing increases network performance by eliminating duplicate message copying in both the host OS and the guest OS. Furthermore, socket-outsourcing significantly enhances inter-VM communication within the same host OS since it enables network packets to bypass the protocol stack in guest OSes. Socket-outsourcing was implemented in two representative operating systems (Linux and NetBSD) and on two virtual machine monitors (Linux KVM and PansyVM). These virtual machine monitors provided support for socket-outsourcing through shard memory, event queues, and VM-specific Remote Procedure Call between a guest OS and a host OS. The experimental results revealed that a guest OS outsourcing the socket layer achieved the same network throughput as a native OS using up to four Gigabit Ethernet links. Moreover, the benchmark results obtained from an N-tier Web application that generated a significant amount of inter-VM communication indicated that socket-outsourcing improved performance by up to 45 percent compared with conventional hosted VM environments. Copyright 2009 acm.
暂无评论