<?xml version="1.0" encoding="UTF-8"?><xml><records><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>17</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Du, Wei</style></author><author><style face="normal" font="default" size="100%">Liao, Yongjun</style></author><author><style face="normal" font="default" size="100%">Tao, Narisu</style></author><author><style face="normal" font="default" size="100%">Geurts, Pierre</style></author><author><style face="normal" font="default" size="100%">Fu, Xiaoming</style></author><author><style face="normal" font="default" size="100%">Leduc, Guy</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Rating Network Paths for Locality-Aware Overlay Construction and Routing</style></title><secondary-title><style face="normal" font="default" size="100%">IEEE/ACM Transactions on Networking</style></secondary-title></titles><keywords><keyword><style  face="normal" font="default" size="100%">matrix factorization</style></keyword><keyword><style  face="normal" font="default" size="100%">network inference</style></keyword><keyword><style  face="normal" font="default" size="100%">rating-based network measurement</style></keyword><keyword><style  face="normal" font="default" size="100%">recommender system</style></keyword></keywords><dates><year><style  face="normal" font="default" size="100%">2015</style></year><pub-dates><date><style  face="normal" font="default" size="100%">10/2015</style></date></pub-dates></dates><volume><style face="normal" font="default" size="100%">23</style></volume><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;p&gt;This paper investigates the rating of network paths,&amp;nbsp;i.e. acquiring quantized measures of path properties such as&amp;nbsp;round-trip time and available bandwidth. Comparing to finegrained&amp;nbsp;measurements, coarse-grained ratings are appealing in&amp;nbsp;that they are not only informative but also cheap to obtain.&lt;/p&gt;&lt;p&gt;Motivated by this insight, we firstly address the scalable&amp;nbsp;acquisition of path ratings by statistical inference. By observing&amp;nbsp;similarities to recommender systems, we examine the applicability&amp;nbsp;of solutions to recommender system and show that our&amp;nbsp;inference problem can be solved by a class of matrix factorization&amp;nbsp;techniques. A technical contribution is an active and progressive&amp;nbsp;inference framework that not only improves the accuracy by&amp;nbsp;selectively measuring more informative paths but also speeds&amp;nbsp;up the convergence for available bandwidth by incorporating its&amp;nbsp;measurement methodology.&lt;/p&gt;&lt;p&gt;Then, we investigate the usability of rating-based network&amp;nbsp;measurement and inference in applications. A case study is&amp;nbsp;performed on whether locality awareness can be achieved for&amp;nbsp;overlay networks of Pastry and BitTorrent using inferred ratings.&lt;/p&gt;&lt;p&gt;We show that such coarse-grained knowledge can improve the&amp;nbsp;performance of peer selection and that finer granularities do not&amp;nbsp;always lead to larger improvements.&lt;/p&gt;</style></abstract><issue><style face="normal" font="default" size="100%">5</style></issue><section><style face="normal" font="default" size="100%">1661</style></section></record><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>17</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Yongjun Liao</style></author><author><style face="normal" font="default" size="100%">Wei Du</style></author><author><style face="normal" font="default" size="100%">Pierre Geurts</style></author><author><style face="normal" font="default" size="100%">Guy Leduc</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">DMFSGD: A decentralized matrix factorization algorithm for network distance prediction</style></title><secondary-title><style face="normal" font="default" size="100%">IEEE/ACM Transactions on Networking</style></secondary-title></titles><keywords><keyword><style  face="normal" font="default" size="100%">matrix completion</style></keyword><keyword><style  face="normal" font="default" size="100%">matrix factorization</style></keyword><keyword><style  face="normal" font="default" size="100%">network distance prediction</style></keyword><keyword><style  face="normal" font="default" size="100%">stochastic gradient descent</style></keyword></keywords><dates><year><style  face="normal" font="default" size="100%">2013</style></year><pub-dates><date><style  face="normal" font="default" size="100%">10/2013</style></date></pub-dates></dates><volume><style face="normal" font="default" size="100%">21</style></volume><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;p&gt;The knowledge of end-to-end network distances is&amp;nbsp;essential to many Internet applications. As active probing of all&amp;nbsp;pairwise distances is infeasible in large-scale networks, a natural&amp;nbsp;idea is to measure a few pairs and to predict the other ones&amp;nbsp;without actually measuring them. This paper formulates the&amp;nbsp;prediction problem as matrix completion where the unknown&amp;nbsp;entries in a pairwise distance matrix constructed from a network&amp;nbsp;are to be predicted. By assuming that the distance matrix has&amp;nbsp;a low-rank characteristics, the problem is solvable by lowrank&amp;nbsp;approximation based on matrix factorization. The new&amp;nbsp;formulation circumvents the well-known drawbacks of existing&amp;nbsp;approaches based on Euclidean embedding.&lt;/p&gt;&lt;p&gt;A new algorithm, so-called Decentralized Matrix Factorization&amp;nbsp;by Stochastic Gradient Descent (DMFSGD), is proposed. By&amp;nbsp;letting network nodes exchange messages with each other, the&amp;nbsp;algorithm is fully decentralized and only requires each node&amp;nbsp;to collect and to process local measurements, with neither&amp;nbsp;explicit matrix constructions nor special nodes such as landmarks&amp;nbsp;and central servers. In addition, we compared comprehensively&amp;nbsp;matrix factorization and Euclidean embedding to demonstrate&amp;nbsp;the suitability of the former on network distance prediction. We&amp;nbsp;further studied the incorporation of a robust loss function and&amp;nbsp;of non-negativity constraints. Extensive experiments on various&amp;nbsp;publicly-available datasets of network delays show not only the&amp;nbsp;scalability and the accuracy of our approach, but also its usability&amp;nbsp;in real Internet applications.&lt;/p&gt;</style></abstract><issue><style face="normal" font="default" size="100%">5</style></issue><section><style face="normal" font="default" size="100%">1511</style></section></record></records></xml>