针对传统K近邻分类器在大规模数据集中存在时间和空间复杂度过高的问题,可采取原型选择的方法进行处理,即从原始数据集中挑选出代表原型(样例)进行K近邻分类而不降低其分类准确率.本文在CURE聚类算法的基础上,针对CURE的噪声点不易确定...针对传统K近邻分类器在大规模数据集中存在时间和空间复杂度过高的问题,可采取原型选择的方法进行处理,即从原始数据集中挑选出代表原型(样例)进行K近邻分类而不降低其分类准确率.本文在CURE聚类算法的基础上,针对CURE的噪声点不易确定及代表点分散性差的特点,利用共享邻居密度度量给出了一种去噪方法和使用最大最小距离选取代表点进行改进,从而提出了一种新的原型选择算法PSCURE (improved prototype selection algorithm based on CURE algorithm).基于UCI数据集进行实验,结果表明:提出的PSCURE原型选择算法与相关原型算法相比,不仅能筛选出较少的原型,而且可获得较高的分类准确率.展开更多
Fast resources discovery and high Quality of Service(QoS) guarantee are the key determinants for efficient Mobile P2P(MP2P) video sharing.In this paper,we propose a CrossLayer and One-Hop Neighbour-Assisted Video Shar...Fast resources discovery and high Quality of Service(QoS) guarantee are the key determinants for efficient Mobile P2P(MP2P) video sharing.In this paper,we propose a CrossLayer and One-Hop Neighbour-Assisted Video Sharing(CNVS) solution in Mobile Ad Hoc Networks(MANETs).By making use of crosslayer approaches to bridge the overlay and the MAC layer and with the help of dissemination assisted by one-hop neighbours,CNVS intelligently builds the resource-centric self-organization node cluster group.In order to meet the QoS requirement,by making use of video resources access cost model,each peer can disconnect a less efficient connection with the original supplier and choose a peer that provides a low access cost as the new supplier.Simulation results also show how CNVS achieves lower average end-to-end delay,less average number of hops for video data delivery,lower routing overhead and packet loss rate,and higher network throughput than another state-of-the-art solution.展开更多
Many territorial animals exhibit reduced aggression towards neighbours. Known as ＂the dear enemy effect＂, this phenomenon has been documented among conspecific animals across a wide range of animal taxa. In theory, ...Many territorial animals exhibit reduced aggression towards neighbours. Known as ＂the dear enemy effect＂, this phenomenon has been documented among conspecific animals across a wide range of animal taxa. In theory, the dear enemy ef- fect can also exist between individuals of different species, particularly when those species compete for shared resources. To date, a heterospecific dear enemy effects has only been documented in ants. In this study, we test for both a conspecific and heterospe- cific dear enemy effect in neotropical rufous-and-white wrens Thryophilus rufalbus. This species competes for resources with banded wrens Thryophilus pleurostictus, a closely related sympatric congener. We used acoustic playback to simulate ruf- ous-and-white wren and banded wren neighbours and non-neighbours at the edges of rufous-and-white wren territories. Ruf- ous-and-white wrens responded more strongly to signals from their own species, demonstrating that resident males discriminate between conspecific and heterospecific rivals. They did not, however, exhibit a conspecific dear enemy effect. Further, they did not exhibit a heterospecific dear enemy effect. This could be due to neighbours and non-neighbours posing similar levels of threat in this system, to the possibility that playback from the edges of the subjects＇ large territories did not simulate a threatening signal, or to other factors. Our study provides the first test of a heterospecific dear enemy effect in vertebrates, and presents a valuable experimental approach for testing for a heterospecific dear enemy effect in other animals [Current Zoology 61 （1）： 23-33, 2015].展开更多
文摘针对传统K近邻分类器在大规模数据集中存在时间和空间复杂度过高的问题,可采取原型选择的方法进行处理,即从原始数据集中挑选出代表原型(样例)进行K近邻分类而不降低其分类准确率.本文在CURE聚类算法的基础上,针对CURE的噪声点不易确定及代表点分散性差的特点,利用共享邻居密度度量给出了一种去噪方法和使用最大最小距离选取代表点进行改进,从而提出了一种新的原型选择算法PSCURE (improved prototype selection algorithm based on CURE algorithm).基于UCI数据集进行实验,结果表明:提出的PSCURE原型选择算法与相关原型算法相比,不仅能筛选出较少的原型,而且可获得较高的分类准确率.
基金supported in part by the National High Technical Research and Development Program of China (863 Program) under Grant No. 2011AA010701the National KeyBasic Research Program of China (973 Program)under Grant No. 2013CB329102+4 种基金the National Natural Science Foundation of China under Grants No. 61001122,No. 61003283,No. 61232017the Natural Science Foundation of Jiangsu Province under Grant No. BK2011171the Natural Science Foundation of Jiangxi,China under Grant No. 20122BAB201042the Fundamental Research Funds for the Central Universities under Grant No. 2012RC0603the Science Foundation Ireland under Grant No. 10/CE/I1855 to Lero-the Irish Software Engineering Research Centre
文摘Fast resources discovery and high Quality of Service(QoS) guarantee are the key determinants for efficient Mobile P2P(MP2P) video sharing.In this paper,we propose a CrossLayer and One-Hop Neighbour-Assisted Video Sharing(CNVS) solution in Mobile Ad Hoc Networks(MANETs).By making use of crosslayer approaches to bridge the overlay and the MAC layer and with the help of dissemination assisted by one-hop neighbours,CNVS intelligently builds the resource-centric self-organization node cluster group.In order to meet the QoS requirement,by making use of video resources access cost model,each peer can disconnect a less efficient connection with the original supplier and choose a peer that provides a low access cost as the new supplier.Simulation results also show how CNVS achieves lower average end-to-end delay,less average number of hops for video data delivery,lower routing overhead and packet loss rate,and higher network throughput than another state-of-the-art solution.
文摘Many territorial animals exhibit reduced aggression towards neighbours. Known as ＂the dear enemy effect＂, this phenomenon has been documented among conspecific animals across a wide range of animal taxa. In theory, the dear enemy ef- fect can also exist between individuals of different species, particularly when those species compete for shared resources. To date, a heterospecific dear enemy effects has only been documented in ants. In this study, we test for both a conspecific and heterospe- cific dear enemy effect in neotropical rufous-and-white wrens Thryophilus rufalbus. This species competes for resources with banded wrens Thryophilus pleurostictus, a closely related sympatric congener. We used acoustic playback to simulate ruf- ous-and-white wren and banded wren neighbours and non-neighbours at the edges of rufous-and-white wren territories. Ruf- ous-and-white wrens responded more strongly to signals from their own species, demonstrating that resident males discriminate between conspecific and heterospecific rivals. They did not, however, exhibit a conspecific dear enemy effect. Further, they did not exhibit a heterospecific dear enemy effect. This could be due to neighbours and non-neighbours posing similar levels of threat in this system, to the possibility that playback from the edges of the subjects＇ large territories did not simulate a threatening signal, or to other factors. Our study provides the first test of a heterospecific dear enemy effect in vertebrates, and presents a valuable experimental approach for testing for a heterospecific dear enemy effect in other animals [Current Zoology 61 （1）： 23-33, 2015].
