Distributed flow algorithms for scalable similarity visualization

We describe simple yet scalable and distributed algorithms for solving the maximum flow problem and its minimum cost flow variant, motivated by problems of interest in objects similarity visualization. We formulate the fundamental problem as a convex-concave saddle point problem. We then show that this problem can be efficiently solved by a first order method or by exploiting faster quasi-Newton steps. Our proposed approach costs at most O(|ε|) per iteration for a graph with |ε| edges. Further, the number of required iterations can be shown to be independent of number of edges for the first order approximation method. We present experimental results in two applications: mosaic generation and color similarity based image layouting. © 2010 IEEE.

彭州市白水河自然保护区生物多样性及种子植物区系研究

对白水河自然保护区进行了生物多样性和植物区系调查，根据获得的生物多样性数据和标本的鉴定结果分析，得到结果和推论如下：1．白水河自然保护区生物多样性的垂直分布格局1.1 植物群落α 多样性随海拔梯度的变化乔木种的丰富度及多样性随海拔上升表现出明显的线性下降趋势。而灌木和草本物种丰富度及多样性随海拔上升表现出抛物线式的下降趋势。乔木种从海拔1400m 的15 种至林线时下降为2 种；灌木和草本植物分别从35 和38 种至山顶时下降为5 种和20 种。乔木物种随海拔升高出现明显的物种替代现象，表明海拔梯度包含了多种环境因子的梯度效应，影响着植物群落的分布与结构及物种多样性。1.2 植物群落β 多样性随海拔梯度的变化海拔2200 m 左右是一个明显的生境转折点。海拔2200 米以下相邻群落的相似性( CJ ) 明显大于海拔2200 m 以上的群落，说明海拔2200 m 以下的群落间共有种多，生境差异较小；而海拔2200 以上的群落则相反，相似性较低。低海拔区由于人为干扰较大，所以群落具有较高的物种丰富度，相邻群落之间的物种替换总量(Cody 指数) 较大。海拔2800 米到海拔3200 米之间因杜鹃群落的影响，物种替换总量(Cody 指数) 略有升高。研究β 多样性沿海拔梯度的变化必须考虑到物种丰富度和群落类型的影响，用不同指数从不同角度能更好地理解β 多样性沿环境梯度的变化。2．白水河自然保护区植物区系性质及起源白水河自然保护区自产种子植物计138 科421 属990 种。本文在科、属的水平上对该保护区植物区系特性进行了较深入的统计和分析。统计表明，温带和热带分布型均占有相当比重，但温带分布型稍占优势；热带、亚热带和温带的科、属多，中国特有属也有相当比例，它们是保护区具有特征意义的类群，其中许多属为古老和残遗成分。结论认为白水河保护区植物区系起源古老，较完好的保存了北极－第三纪古植物群。We have collected the specimens and gotten the biodiversity datas of BaiShuihe Nature Reserve in Peng Zhou，analysed the datas，the results as follows：1. Diversity of the plant community along altitudinal gradient1.1 α diversity of the plant communities along altitudinal gradientFrom 1400m to 3900m at Baishuihe Nature Reserve, 52 plots were investigatedwith an interval 100m in altitude； α diversity and β diversity of plant communitiesand their variety along altitudinal gradient were studied. The results showed that indifferent successional layers of trees, richness and diversity decreased linearly withthe increase of altitude. But shrub and herb layers don’t decrease linearly with theincrease of altitude. Tree species decreased from 15 species at 1400m so only 2species at timberline. Shrub and herb species decreased from 35 and 38 species at2000m to 5 and 20 species at 3800m respectively. Tree species are replacedobviously with the increase of altitude； It shows that altitude includes manyenvironmental facts, which infect the distribution, structure and diversity of plant population.1.2 The variety of βdiversity along altitudinal gradient.The entironment changed obviously near 2200m according to our research. Forsimilarity(CJ) between neighboring plots above 2200m is larger than wich below it.It shows that below 2200m，the neighboring plots has more same species，and thehabitats of neighboring plots has more similarity. Above 2200 is the other way round.The plant communites have higher species richness and species turnoverlargestly between neighboring plots because of the disturbance from humanity at lowaltitude. Between 2800m and 3200m species turnover not so obviously because ofmore Rhododendron live there.We should think over species and the types of plant communities effect thevariety of β diversity along altitude gradient. Use more biodiversity indexs and fromseveral aspects to understand the variety of β diversity along altitudinal gradient.2.Origin and characteristic features of Bai Shuihe Nature Reserve990 species of wild seed plants (belonging to 421 generas in 138 families) inthe floristic region of Bai Shuihe Nature Reserve were reported here. The statisticsand comparatively intensive analysis at generic and familiar levelss. Based on thestatistics，the results show that both temperate and tropical distribution types areacounted for considerable proportion of the total，but formal is a little moreimportant than the later. The North Temperate and E.Asia-N.America disjunctedpatterns are more concentrated in this area. These may be considered as thecharacteristic features of Bai Shuihe Nature Reserve flora, while many of them arearchaic and relic elements. According to above data，the floristic region of BaiShuihe Nature Reserve may be considered as a typical region in Chinese flora. Also，the flora of Bai Shuihe Nature Reserve are originated since ancient time as a wellconserved Arctic-Tertiar flora.

四川九顶山西坡及龙肘山维管植物区系和植物多样性研究

横断山地区是一个十分自然的植物区系地区，在中国植物区系分区中是作为泛北极植物区中国-喜马拉雅亚区中的一个地区，其种子植物区系具有丰富的科、属、种，地理成分复杂，特有现象和替代现象明显。该地区作为植物区系和生物多样性的研究热点地区，长期以来极受中外植物学家关注。横断山脉东缘是中国-喜马拉雅和中国-日本植物区系的交汇过渡区域，北部的岷江流域以及南部的金沙江流域，孕育了该区丰富的物种资源和植被资源。而岷江干热河谷和金沙江干热河谷的相似性和相关性，更为该区的植物区系和生物多样性南北的对比研究提供了有利的条件。 本研究选择的九顶山西坡和龙肘山分别位于横断山区北部和南部，九顶山属岷江流域而龙肘山属金沙江流域。本研究结合植物区系研究和生物多样性研究，对该区的植物资源进行调查。通过样带调查和样线踏查结合，大量详实的野外样方调查和标本采集，进行传统的区系研究和生物多样性研究。研究该区物种多样性的海拔梯度格局及其潜在的影响影子，并利用新的区系评估质量方法对九顶山西坡的植物区系质量进行定量的研究，以期能更为深刻的理解该区的植物资源，为该区的资源保护和利用提供合理可行的建议。主要研究结论如下： 1）九顶山西坡植物区系的性质和特点 经鉴定和统计，九顶山西坡共有1707 种维管植物，分属617 属和140 科,其中种子植物1616 种，分属572 属117 科。就科的分布区成分构成而言，该区系的热带成分与温带成分相当，热带成分略占优势，表明九顶山西坡的植物区系与热带植物区系有较强的联系。但是，在九顶山西坡属的分布区类型所占的比例上，温带成分远远超过了热带成分，本区的种子植物分布表现出明显的温带性质。且温带分布类型的许多物种组成了九顶山西坡植被的建群种和优势种，是本区系最重要的成分，充分体现了本区系的温带性质。 2）九顶山西坡不同植被带的生物多样性海拔梯度格局 基于对土门-断头崖、茶山-九顶山、雁门沟-光光山三条垂直植被样带的调查，我们发现九顶山西坡的生物多样性沿海拔梯度的变化呈现出一定的规律性，不同样带之间有一定差异。就三条样带的物种组成相似性来看，虽然土门－断头崖样带属于涪江水系，而茶山－九顶山样带和雁门沟－光光山样带属于岷江水系，但不同水系对该区物种组成的影响并不明显。三条样带中，草本层物种丰富度均远远大于灌木层和乔木层，而以乔木层物种丰富度最低；α－多样性指数随着海拔梯度的变化在土门-断头崖样带中呈现单一下降趋势，在茶山-九顶山样带表现为双峰模型，而在雁门沟-光光山样带则表现为不显著波动变化；均匀度指数在土门-断头崖样带呈现出单一下降的趋势，在雁门沟-光光山样带表现为凹形曲线，而在茶山-九顶山样带却无明显的变化规律。β－多样性指数在土门-断头崖样带和茶山-九顶山样带呈现出明显的波动状态，植被类型替代现象明显；而在雁门沟-光光山样带却并未有十分显著的转折点，因其水平植被带受到干扰，同海拔替代现象不显著。 3）九顶山西坡维管植物丰富度的海拔梯度格局 我们考察了九顶山西坡和两条垂直样带（土门－断头崖和雁门沟－光光山样带）的不同分类等级（包括科、属、种）和不同生活型物种（乔木、灌木、禾草、蕨类和其它草本）的丰富度沿着海拔梯度的分布。结果发现，物种的丰富度海拔梯度格局具有不同的模式，单一下降和中间膨胀格局依然是其主流。不同生活型的物种具有不同的丰富度格局，但是对于环境需求相似的类型具有较相似的丰富度格局。不同的丰富度格局可能由多因素导致，包括：气候，海拔跨度，面积，人为干扰等等。 4）九顶山西坡区系质量评估 我们尝试使用传统的区系质量评估方法对九顶山西坡的区系质量进行评估，并尝试使用一种新的区系质量评估体系对该区的区系进行评价。在九顶山西坡随着海拔梯度的上升，平均保守性系数在各条植被带中均呈现出逐渐上升的趋势。区系质量指数随着海拔的升高都表现为双峰模型，在植被交错区区系质量指数相对较高，而在海拔的两极，区系质量指数都很低。大部分地区使用新方法计算所得的加权平均保守性系数和区系质量指数都比传统方法计算的平均保收性系数和区系质量指数要高，说明在九顶山西坡的三条样带中，大部分地区都是那些保守性系数较高的物种占据优势，同时也表明九顶山西坡具有很高质量的区系和自然植被。 5）龙肘山种子植物区系的性质和特点 龙肘山种子植物区系的物种较为丰富，共有154 科，544 属，1156 种。科的优势十分明显，单种属和寡种属数量众多，说明本区系植物成分较为复杂、起源古老、物种多样性指数较高。地理成分复杂，分布类型多样，其中热带成分在总数量上高于温带成分，但是许多温带成分的属是该区植被的重要建群类群和优势类群，表现出明显的亚热带性质。 6）龙肘山生物多样性的现状和特点 在海拔梯度上，龙肘山地区无论是科、属、种的数量，还是不同等级分类单元之间的数量比，均呈现先升后降的趋势，并在中海拔地区达到峰值。物种多样性指数从总体上来说变化幅度不大，略有先升后降的趋势，在中海拔梯度物种多样性最高。乔、灌、草三层的多样性指数表现出乔木层<灌木层<草本层的特征；乔木层均匀度的变化很大，而灌木层和草本层均匀度的变化较小；灌木层均匀度的波动又强于草本层。β－多样性指数呈现单峰模式，中海拔地区最高。就龙肘山东、西坡物种多样性相比较而言，两者虽然在数值上交替上升，但是却体现出了较为一致的趋势，但西坡因受到干热河谷气候的影响，其平均气温要高于东坡，导致了东坡植物群落和物种的分布比西坡要低。在区系成分构成上，低山区的相同海拔段，西坡的热带亚热带成分所占的比例要比东坡高，这是因为西坡的平均气温比东坡稍高，导致了热带、亚热带物种分布更多。而随着海拔的上升，东、西两坡的气候、土壤等条件趋于一致，其植物区系成分的构成格局也趋于一致。 The Hengduan Mountain region is a very natural floristic region; it belongs toChina-Himalaya sub-region of Holarctic region in floristic subarea of China. The flora in this areais rich in family, genus and species; has a very complex composition of geographical elements;especially with high richness of endemic species and obvious substitution phenomenon. Thisregion as a hot-spot area of floristic and biodiversity, has fascinated biologists in the world for along time. The eastern range of Hengduan Mountain is the transition zone of China – Himalayaforest sub-region and China-Japan forest sub-region in floristic. The water systems are quitedifferent, Minjiang River in the north and Jishajiang River in the south grow quit different but alsoabundant plant species and vegetation resources. The similarity and correlativity of Minjiang River dry valleys and Jinshajiang River dry valleys have provided advantageous condition tocontrast flora and biodiversity between north and south. In the present study, the Jiuding Mountainlies in the north of Hengduan Mountain and belongs to Minjiang River, and the LongzhouMountain lies in the south of Hengduan Mountain and belongs to Jinshajiang River. In our study, we combined the methods of floristic research and biodiversity investigation toexplore the resources of plant species and vegetations; sampled with transects along the altitudinalgradients and also with transverse straps with similar elevation; collected the vascular plant specimen with sampling plots of ecology. We explored the plant species richness patterns alongaltitudinal gradients and discussed the underlying factors aroused these patterns; and used a novelmethod to assess the quality of Jiuding Mountain’s flora. All for a deeper comprehension of the plant recourses of this region; and provided feasible and reasonable suggestion for the protectionof resources. The results were as follows: 1 The characteristic of the flora of the west slope of Jiuding Mountain We had collected 1707 species of vascular plants belonging to 617 genera in 140 families inthe west slope of Jiuding Mountain，in which included 1616 seed plant species belonging to 572genera and 117 families. As for the composition of the areal types of the Families of seed plants,tropic components and temperate components are well-balanced, and percentage of tropicscomponents is higher than that of temperate ones for a litter bit. This shows the flora in the westslope of Jiuding Mountain has strong relationship with the tropic flora. But for the composition ofthe areal types of genera, temperate components have far exceeded the tropics ones, indicated thewhole flora with a conspicuous temperate character. Temperate components possess maximumproportion in the west slope of Jiuding Mountain, and many of them belong to constructivespecies and dominant species in the vegetation, are most important components in JiudingMountain’s Flora, also have embodied the temperate character of this area sufficiently. 2 Biodiversity patterns along altitudinal gradients in different vegetation transects in the westslope of Jiuding Mountain Based on the investigation of three vegetation transects (including Tumen-Duantouya transect,Chashan-Jiudingshan transect and Yanmengou-Guangguangshan Transect) in the west slope ofJiuding Mountain, we found the change of biodiversity along the altitude gradients displayedcertain regularity, but have differences among different transects. The three transects belong todifferent water systems; the Tumen-Duantouya transect belongs to Fujiang River, and the othertwo belong to Minjiang River. From the similarity of species compositions of different transects,we found different water system didn’t show obvious impact on the species composition. In all thethree transects, the species richness of herb layer was remarkably higher than shrub and tree layer,and the species richness of tree layer was the lowest one. With the increasing of the altitude, theline of α-diversity was monotonically decreasing curve in Tumen-Duantouya transect, andbimodal curve in Chashan-Jiudingshan transect, but in Yanmengou-Guangguangshan transectshowed a wave-like curve although not very obvious. Species evenness showed monotonicallydecreasing trends in Tumen-Duantouya transect, and very low at mid-altitude in Yanmengou-Guangguangshan transect, but in Chashan-Jiudingshan transect changed irregularly. Changes inβ-diversity corresponded with the transition of vegetation in the Tumen-Duantouya transect andChashan-Jiudingshan transect, and the curve of β-diversity along altitude had obvious turningpoint; but in Yanmengou-Guangguangshan transect had no obvious turning point, and thesubstitution phenomenon was not obvious, transverse vegetation straps distributed interlaced. 3 Richness patterns of vascular plant species along altitude in the west slope of Jiuding Mountain Direct gradient analysis and regression methods were used to describe the species richnesspatterns along the altitudinal for Mt. Jiuding, as well as separately for Tumen-Duantouya Transectand Yanmengou-Guangguangshan Transect. Altitudinal gradient of diversity of units at differenttaxonomic level (including Family, Genus and Species) and at different life form (including tree,shrub, pteridophyte, grass and other herb) were tested to find differences among the richnesspattern. We found altitudinal richness also shows different patterns, and both monotonicallydecreasing pattern and hump-shaped pattern can be founded in vascular species richness. Speciesin different life forms show different altitudinal patterns, but those species with similarrequirements to environmental conditions show similar richness patterns along altitudinalgradients. Different richness patterns can be aroused by different climate, different altitudinal span,area factor, anthropogenic factor and so on. 4 Floristic quality assessments in the west slope of Jiuding Mountain We used both the conventional method broadly adopted in the USA and the new one toassess the floristic quality in the west slope of Jiuding Mountain. The Mean Coefficient ofConservatism (MC) had the trend of increment along the altitudinal gradients. The FloristicQuality Index (FQI) was a bimodal curve with increasing of elevation; FQI got maximum valuesin the transition zones of different vegetations in the middle altitude, and had very low values atthe two end of elevation. In most areas of the west slope of the Jiuding Mountain, the resultscalculated using the new methods were higher than those using the conventional method. Thisindicated the dominant species of the communities had very high coefficients of conservatism inmost areas of Jiuding Mountain, and the communities are relatively kept pristine and the habitats very integrative. 5 The characteristic of the flora of Longzhou Mountain The flora of Longzhou Mountain has very abundant in species composition; there are about1156 species of seed plants belonging to 544 genera in 154 families. In which, twelve families with more than 20 species include totally 232 genera and 532 species, and form the majority of itsflora. The origin of its flora is old, monospecific genera and oligotypic genera amounts to 510 innumber, which constitute 93.75% of total number of genera. The geographical components arevarious in Longzhou Mountain, the majority of flora are temperate and pantropic ones. The tropiccomponents overtopped temperate components on genera quantity, but many temperatecomponents belong to constructive species and dominant species in the vegetation, and the wholeflora shows an obvious subtropical character. 6 Current situation and characteristic of biodiversity in Longzhou Mountain With the increasing of altitude, the number of species, genus, family and the ratios ofdifferent taxonomic levels all displayed a trend of descending after rising first, and peaked atmiddle height area. The change of α-diversity was not very acutely, with the trend of descendingafter rising first in some degree, the middle height area had highest α-diversity. As studying thetree layer, shrub layer and herb layer respectively, the Shannon-Wiener index was in followingorder: tree layer < shrub layer < herb layer; the change of evenness was more complicatedly thanthat of diversity, the tree layer changed acutely, but the shrub layer and herb layer fluctuatedsmoothly. Changes in β-diversity also showed the trend of descending after rising first. TheJaccard index and Cody index all peaked at the middle height forest area. As for the comparison ofplant diversity and evenness between the west and east slope, the numerical values ascendedalternatively, but the trend of changing was similar. The distribution of similar plant communitiesand species in east slope were lower than the west slope for the influence of Jinsha River DryValley. As for the composition of different floristic components, in lower altitude area of westslope, the tropic and sub-tropic plants had higher ratio than east slope’s and even could be equal tothe temperate plants. With the increasing of elevation, the floristic composition become morelikely between the east and west slope and temperate plants dominated the flora.

Graphical representation and the similarity of DNA primary sequences

In this article, graphical representations of DNA primary sequences were generated. Topological indices and molecular connectivity indices were calculated and used for the comparison of similarities among eight different DNA segments. The satisfactory results were achieved by this analysis.

Local similarity in organic crystals and the non-uniqueness of X-ray powder patterns

Two new concepts for molecular solids, 'local similarity' and 'boundary-preserving isometry', are defined mathematically and a theorem which relates these concepts is formulated. 'Locally similar' solids possess an identical short-range structure and a 'boundary-preserving isometry' is a new mathematical operation on a finite region of a solid that transforms mathematically a given solid to a locally similar one. It is shown further that the existence of such a 'boundary-preserving isometry' in a given solid has infinitely many 'locally similar' solids as a consequence. Chemical implications, referring to the similarity of X-ray powder patterns and patent registration, are discussed as well. These theoretical concepts, which are first introduced in a schematic manner, are proved to exist in nature by the elucidation of the crystal structure of some diketopyrrolopyrrole (DPP) derivatives with surprisingly similar powder patterns. Although the available powder patterns were not indexable, the underlying crystals could be elucidated by using the new technique of ab initio prediction of possible polymorphs and a subsequent Rietveld refinement. Further ab initio packing calculations on other molecules reveal that 'local crystal similarity' is not restricted to DPP derivatives and should also be exhibited by other molecules such as quinacridones. The 'boundary-preserving isometry' is presented as a predictive tool for crystal engineering purposes and attempts to detect it in crystals of the Cambridge Structural Database (CSD) are reported.

Atom space distance and computing of molecular similarity

A mathematical model on computation of molecular similarity was suggested, The algorithmic techniques for measuring the degree of similarity between pairs of three-dimensional chemical molecules was represented by modified interatomic distance matrices. Current work was carried out on Indigo 2 work station with Sybyl software. Four groups of molecules were used to compute the molecules similarity to testing the mathematical model with satisfactory results.

Chemical environment code and measure of molecular similarity

A new scheme for the code of chemical environments of compounds is described in this paper, and three molecular similarity methods have been used to select nearest neighbors from four different types of probe compounds. One of the methods is based on the C-13 NMR spectra. The second method is based on the code of chemical environments and molecular topological index A(x). The third approach, i.e. the Tanimoto coefficient, is also based on the code of chemical environments, but not to use the topological index. Five nearest neighbors for each probe compound using these three molecular similarity methods were determined and taken from the database of 7309 structures. The results indicate that the scheme of the chemical environment code and the method for similarity measure of intermolecules suggested in this study are reasonable. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.

Code for chemical environment and computation of molecular similarity

A new code for chemical environment and an empirical mathematical pattern Sa(m) on computation of molecular similarity were suggested. Seven molecules which referred to as the probe compounds and the nearest neighbors of each probe structure were determined by the methods of Sa(m) and Tanimoto, The results show an intuitive notion of chemical similarity.

The floristic of the family Gramineae of Karakorum and Kunlun area, China

There are 47 genera and 161 species of Gramineae except the cultivated species in the area of the Karakorum and Kunlun Mountains. The results of research on the distribution of the genera and species of Gramineae in the Karakorum and Kunlun Mountains show that (1) The Gramineae mainly contains elements of North Temperate, rich Old Word Temperate and other Temperate. It is obvious that the floristic nature of Gramineae in the Karakorum and Kunlun Mountains is the North Temperate; (2) All Pantropic genera can stretch to the Temperate Zone in this region, which all parts of the Pantropic type are the Temperate nature to a certain degree. For example, Erianthus ravennae from mediterranean to the Karakorum and Kunlun Mountains through the Central Asia; (3) As most genera of Grasses are the type of Temperate and the Frigid Zone, they have distinct floristic characteristics of mountainous and plateau flora such as Orinus, Alopecurus, Elymus, Trisetum, Littledalea, Elytrigia, Stephanachne and Paracolpodium etc. All of these indicate adaptive phenomenon of alpine specialization and cold-xerophilization on Grasses in this area; (4) Endemic genus of Gramineae is absent due to its nature and history and the endemic species are also rare in the Karakorum and Kunlun Mountains. Most of the genera with one or fewer species have originated from its relative and widespread genera, such as Ptilagrostis from Stipa, Timouria from Achnatherum, and so on; (5) Flora of the Karakorum and Kunlun Mountains is most closely related to the flora of Tibet, and is also extensively to its adjacent areas.

Homology Induction: the use of machine learning to improve sequence similarity searches

Karwath, A. King, R. Homology induction: the use of machine learning to improve sequence similarity searches. BMC Bioinformatics. 23rd April 2002. 3:11 Additional File Describes the title organims species declaration in one string [http://www.biomedcentral.com/content/supplementary/1471- 2105-3-11-S1.doc] Sponsorship: Andreas Karwath and Ross D. King were supported by the EPSRC grant GR/L62849.

Semi-Supervised OWA Aggregation for Link-Based Similarity Evaluation and Alias Detection

T.Boongoen and Q. Shen. Semi-Supervised OWA Aggregation for Link-Based Similarity Evaluation and Alias Detection. Proceedings of the 18th International Conference on Fuzzy Systems (FUZZ-IEEE'09), pp. 288-293, 2009. Sponsorship: EPSRC

BoostMap: A Method for Efficient Approximate Similarity Rankings

This paper introduces BoostMap, a method that can significantly reduce retrieval time in image and video database systems that employ computationally expensive distance measures, metric or non-metric. Database and query objects are embedded into a Euclidean space, in which similarities can be rapidly measured using a weighted Manhattan distance. Embedding construction is formulated as a machine learning task, where AdaBoost is used to combine many simple, 1D embeddings into a multidimensional embedding that preserves a significant amount of the proximity structure in the original space. Performance is evaluated in a hand pose estimation system, and a dynamic gesture recognition system, where the proposed method is used to retrieve approximate nearest neighbors under expensive image and video similarity measures. In both systems, BoostMap significantly increases efficiency, with minimal losses in accuracy. Moreover, the experiments indicate that BoostMap compares favorably with existing embedding methods that have been employed in computer vision and database applications, i.e., FastMap and Bourgain embeddings.

Explaining World Wide Web Traffic Self-Similarity

Recently the notion of self-similarity has been shown to apply to wide-area and local-area network traffic. In this paper we examine the mechanisms that give rise to self-similar network traffic. We present an explanation for traffic self-similarity by using a particular subset of wide area traffic: traffic due to the World Wide Web (WWW). Using an extensive set of traces of actual user executions of NCSA Mosaic, reflecting over half a million requests for WWW documents, we show evidence that WWW traffic is self-similar. Then we show that the self-similarity in such traffic can be explained based on the underlying distributions of WWW document sizes, the effects of caching and user preference in file transfer, the effect of user "think time", and the superimposition of many such transfers in a local area network. To do this we rely on empirically measured distributions both from our traces and from data independently collected at over thirty WWW sites.