Abstract

Abstract
We present a generalization of Widrow's adaptive linear combiner with an adaptive recursive memory. Expressions for memory depth and resolution are derived. The LMS procedure is extended to adapt the memory depth and resolution so as to match the signal characteristics. The particular memory structure, gamma memory, was originally developed as part of a neural net model for temporal processing.

Abstract
The authors present the concept of using a high-speed data communication switch as a backbone network concentrator, with low packet delay, in order to accommodate internet and high-demand traffic. Also presented is the front-end processor concept, namely the implementation of an Ethernet front-end processor to connect Ethernet stations and LANs. Another front-end processor is currently being developed for the Fiber Distributed Data Interface (FDDI), because most of the FDDI applications will be LAN interconnection and direct connection of mainframes, high-speed servers and workstations.

Abstract

Abstract
An interactive software tool for epileptiform spike detection and focus localization is presented. The method for focus localization uses a multistrategy approach based on different and independent computerized methods. Spike detection is performed using an algorithm based on a set of parameters that characterize the waveforms in the time domain. A number of independent techniques is integrated in a Microsoft Windows based environment to select artifact-free electroencephalogram segments. With this software, the clinician may adjust and validate the automatic spike detector as well as combine and integrate a set of independent techniques to reinforce each other in order to achieve a reliable focal indication.

Abstract
Operation problems that result from MV distribution networks with local asynchronous generation sources connected are examined. In order to analyze steady-state operating situations, a load-flow calculation must be performed. The presence of asynchronous generators imposes the adoption of a consistent bus modelization using a new bus type, the PX bus, where P is the active generated power and X is a nonlinear magnetizing reactance. This approach introduces some modifications in the conventional load-flow algorithms, thereby increasing the total number of iterations.