基于邻域彩色变化矢量场的图像边缘检测技术研究*

首先进行了边缘检测系统结构设计,建立了图像邻域彩色变化矢量场的数理模型,提出了用图像邻域彩色变化方向锐度描述图像边缘,进而应用模糊聚类自适应检测边缘.实验表明:与基于梯度的边缘检测技术相比,该方法在噪声抑制以及边缘准确定位上均取得了好的效果,是一种应用广泛的优秀边缘检测算法.     

基于改进K-means聚类的配电网电压暂降频次估计法

传统考虑保护动作特性的电压暂降频次估计法需要获取详尽的保护配置信息,然而配电网保护配置多样,在不同因素如过渡电阻、运行方式、故障类型等的影响下,阶段式保护各级保护区可能产生较大变化,采用传统方法对电压暂降持续时间进行评估可能会产生较大误差。文中提出一种基于改进K-means聚类的配电网电压暂降频次估计方法,在未知线路保护配置基础上,基于电压暂降历史监测数据与保护动作信息,采用改进K-means聚类算法,对电压暂降幅值-持续时间进行聚类分析,推断线路保护配置情况,计算保护动作时间与保护动作电压。根据计算结果,在考虑不同故障类型、不同运行方式及不同过渡阻抗的情况下进行配电网电压暂降频次估计。在IEEE RBTS-6母线测试系统的母线5配电网中进行仿真,验证了文中方法的有效性和优越性。     

Determination of temporal information granules to improve forecasting in fuzzy time series

Partitioning the universe of discourse and determining intervals containing useful temporal information and coming with better interpretability are critical for forecasting in fuzzy time series. In the existing literature, researchers seldom consider the effect of time variable when they partition the universe of discourse. As a result, and there is a lack of interpretability of the resulting temporal intervals. In this paper, we take the temporal information into account to partition the universe of discourse into intervals with unequal length. As a result, the performance improves forecasting quality. First, time variable is involved in partitioning the universe through Gath–Geva clustering-based time series segmentation and obtain the prototypes of data, then determine suitable intervals according to the prototypes by means of information granules. An effective method of partitioning and determining intervals is proposed. We show that these intervals carry well-defined semantics. To verify the effectiveness of the approach, we apply the proposed method to forecast enrollment of students of Alabama University and the Taiwan Stock Exchange Capitalization Weighted Stock Index. The experimental results show that the partitioning with temporal information can greatly improve accuracy of forecasting. Furthermore, the proposed method is not sensitive to its parameters.     

Leveraging clustering approaches to solve the gray-sheep users problem in recommender systems

Recommender systems apply data mining and machine learning techniques for filtering unseen information and can predict whether a user would like a given item. This paper focuses on gray-sheep users problem responsible for the increased error rate in collaborative filtering based recommender systems. This paper makes the following contributions: we show that (1) the presence of gray-sheep users can affect the performance – accuracy and coverage – of the collaborative filtering based algorithms, depending on the data sparsity and distribution; (2) gray-sheep users can be identified using clustering algorithms in offline fashion, where the similarity threshold to isolate these users from the rest of community can be found empirically. We propose various improved centroid selection approaches and distance measures for the K-means clustering algorithm; (3) content-based profile of gray-sheep users can be used for making accurate recommendations. We offer a hybrid recommendation algorithm to make reliable recommendations for gray-sheep users. To the best of our knowledge, this is the first attempt to propose a formal solution for gray-sheep users problem. By extensive experimental results on two different datasets (MovieLens and community of movie fans in the FilmTrust website), we showed that the proposed approach reduces the recommendation error rate for the gray-sheep users while maintaining reasonable computational performance.     

Impact of class clustering in a multiclass FCFS queue with order-dependent service times

In multi-class queueing systems, customers of different classes can enter the system. When studying such systems, it is traditionally assumed that the different classes of customers occur randomly and independently in the arrival stream of customers in the system. This is often in contrast to the actual situation. Therefore, we study a multi-class system with so-called class clustering in the customer arrival stream, i.e., (Markovian) correlation occurs in the classes of consecutive customers. The system under investigation consists of one server that is able to serve two classes of customers. In addition, the service-time distribution of a customer depends on the equality or non-equality of its class with the class of the previous customer. This latter feature occurs frequently in practice. For instance, execution of the same task again can lead to both faster or slower processing times. The first case can occur when the execution of a different task entails resetting a machine, or loading new data, et cetera. The opposite situation appears, for instance, when execution of the same task requires postprocessing (such as cooling down or reinitialization of a machine). We deduce the probability generating function (pgf) of the system content, from which we can extract various performance measures, among which the mean values of the system content and the customer delay. We demonstrate that class clustering has a tremendous impact on the system performance, which highlights the necessity to include it in the performance assessment of any system in which it occurs.     

Detecting similarities in virtual machine behavior for cloud monitoring using smoothed histograms

The growing size and complexity of cloud systems determine scalability issues for resource monitoring and management. While most existing solutions consider each Virtual Machine (VM) as a black box with independent characteristics, we embrace a new perspective where VMs with similar behaviors in terms of resource usage are clustered together. We argue that this new approach has the potential to address scalability issues in cloud monitoring and management. In this paper, we propose a technique to cluster VMs starting from the usage of multiple resources, assuming no knowledge of the services executed on them. This innovative technique models VMs behavior exploiting the probability histogram of their resources usage, and performs smoothing-based noise reduction and selection of the most relevant information to consider for the clustering process. Through extensive evaluation, we show that our proposal achieves high and stable performance in terms of automatic VM clustering, and can reduce the monitoring requirements of cloud systems.     

A Framework for Interactive t-SNE Clustering

In this paper, we describe our progress in creating the framework for an interactive application that allows humans to actively participate in a t-SNE clustering process. t-SNE (t-Distributed Stochastic Neighbor Embedding) is a dimensionality reduction technique that maps high dimensional data sets to lower dimensions that can then be visualized for human interpretation. By prompting users to monitor outlying points during the t-SNE clustering process, we hypothesize that users may be able to make clustering faster and more accurate than purely algorithmic methods. Further research would test these hypotheses directly. We would also attempt to decrease the lag time between the various components of our application and develop an intuitive approach for humans to aid in clustering unlabeled data. Research into human assisted clustering can combine the strengths of both humans and computer programs to improve the results of data analysis.     

A multi-document summarization system based on statistics and linguistic treatment

The massive quantity of data available today in the Internet has reached such a huge volume that it has become humanly unfeasible to efficiently sieve useful information from it. One solution to this problem is offered by using text summarization techniques. Text summarization, the process of automatically creating a shorter version of one or more text documents, is an important way of finding relevant information in large text libraries or in the Internet. This paper presents a multi-document summarization system that concisely extracts the main aspects of a set of documents, trying to avoid the typical problems of this type of summarization: information redundancy and diversity. Such a purpose is achieved through a new sentence clustering algorithm based on a graph model that makes use of statistic similarities and linguistic treatment. The DUC 2002 dataset was used to assess the performance of the proposed system, surpassing DUC competitors by a 50% margin of f-measure, in the best case.