Search

In the field of signal processing, some problems that are difficult to process using linear theory can be solved by nonlinear processing. As for estimation problems when an unknown system is nonlinear, real-time processing is difficult because of the huge amount of calculations required to find the optimum solution when conventional nonlinear process methods are used. On the other hand, research on neural networks that have nonlinear input-output relationships has attracted attention. These have been used in various areas such as pattern recognition and nonlinear system estimation. In this paper, a method of designing parallel recurrent neural digital filters by introducing several small, recurrent neural networks is proposed for nonlinear adaptive digital filters, which have many parameters and need real-time processing. The proposed method gives better results than conventional methods based on linear and nonlinear theory. Furthermore, learning efficiency can be improved by the proposed method because parallel learning is performed. © 1997 Scripta Technica, Inc. Electron Comm Jpn Pt 3, 80(12): 91–101, 1997 [ABSTRACT FROM AUTHOR]

This paper presents a scalable and adaptive decentralized metadata lookup scheme for ultralarge-scale file systems (more than Petabytes or even Exabytes). Our scheme logically organizes metadata servers (MDSs) into a multilayered query hierarchy and exploits grouped Bloom filters to efficiently route metadata requests to desired MDSs through the hierarchy. This metadata lookup scheme can be executed at the network or memory speed, without being bounded by the performance of slow disks. An effective workload balance method is also developed in this paper for server reconfigurations. This scheme is evaluated through extensive trace-driven simulations and a prototype implementation in Linux. Experimental results show that this scheme can significantly improve metadata management scalability and query efficiency in ultralarge-scale storage systems. [ABSTRACT FROM AUTHOR]