Clustering is the process of discovering groups within multidimensional data, based on similarities, with a minimal, if any, knowledge of their structure. Distributed data clustering is a recent approach to deal with ...
详细信息
Clustering is the process of discovering groups within multidimensional data, based on similarities, with a minimal, if any, knowledge of their structure. Distributed data clustering is a recent approach to deal with geographically distributed databases, since traditional clustering methods require centering all databases in a single dataset. Moreover, current privacy requirements in distributed databases demand algorithms with the ability to process clustering securely. Among the unsupervised neural network models, the self-organizing map (SOM) plays a major role. SOM features include information compression while trying to preserve the topological and metric relationship of the primary data space. This paper presents a strategy for efficient cluster analysis in geographically distributed databases using SOM networks. Local datasets relative to database vertical partitions are applied to distinct maps in order to obtain partial views of the existing clusters. Units of each local map are chosen to represent original data and are sent to a central site, which performs a fusion of the partial results. Experimental results are presented for different datasets.
This paper presents the implementation of ARQ-PROP II, a limited-depth propositional reasoner, via the compilation of its specification into an exact formulation using the satyrus platform. satyrus' compiler takes...
详细信息
Reference architectures are the basis for application instantiation in both Domain engineering and Product Line contexts. They are created based on domain requirements, commonalities, and variability. Considering that...
详细信息
In this paper, we first provide a new theoretical understanding of the Evidence Pre-propagated Importance Sampling algorithm (EPIS-BN) (Yuan & Druzdzel 2003;2006b) and show that its importance function minimizes t...
详细信息
Some real problems are more naturally modeled by hybrid Bayesian networks that consist of mixtures of continuous and discrete variables with their interactions described by equations and continuous probability distrib...
详细信息
Some real problems are more naturally modeled by hybrid Bayesian networks that consist of mixtures of continuous and discrete variables with their interactions described by equations and continuous probability distributions. However, inference in such general hybrid models is hard. Therefore, existing approaches either only deal with special instances, such as Conditional Linear Gaussians (CLGs), or approximate a general model with a restricted version and then perform inference on the simpler model. However, results thus obtained highly depend on the quality of the approximations. This paper describes an importance sampling-based algorithm that directly deals with hybrid Bayesian networks constructed in the most general settings and guarantees to converge to the correct answers given enough time.
Knowledge elicitation is difficult for expert systems that are based on probability theory. The elicitation of probabilities for a probabilistic model requires a lot of time and interaction between the knowledge engin...
详细信息
The establishment of the Women in science and engineering (WiSE) program represents the serious commitment of the University of Southern California to address the under-representation of women in science and engineeri...
详细信息
In order to assist driver's vision, a real-time recognition system for traffic signs is proposed. After detecting sign candidates, biologically inspired opponent-color filters are used to extract symbol parts of s...
详细信息
ISBN:
(纸本)9781617387777
In order to assist driver's vision, a real-time recognition system for traffic signs is proposed. After detecting sign candidates, biologically inspired opponent-color filters are used to extract symbol parts of signs. After normalizing the size of symbol, structural features are calculated to identify the sign. 5572 segmented images are used to design the algorithm. In a real-time system, the same sign in a sequence of frames is tracked, and a majority vote is used to integrate the recognition results. For test data, 93.8% recall rate and 99.3% precision rate could be attained. In-vehicle experiment also showed high recall and precision rates.
In scientific and engineering scenarios we can notice the predominance of explicit knowledge being manipulated and distributed, which makes recommender systems very useful in this environment. But along with a knowled...
详细信息
ISBN:
(纸本)1424409624
In scientific and engineering scenarios we can notice the predominance of explicit knowledge being manipulated and distributed, which makes recommender systems very useful in this environment. But along with a knowledge management approach this kind of system can support the organization in better identifying competences, help engage users in a continuous and dynamic knowledge exchange, and customize knowledge dissemination as much as possible. In this work we detail a collaborative recommender system which is used in a Knowledge Management Environment for Scientific and engineering contexts; we show how this approach can be aimed at a KM process and how this approach can deal with other kinds of knowledge used in research centers and universities.
暂无评论