Abstract: The article deals with a WAN switch design based on a Feed-forward neural network, specifically for the Feed-forward Back-propagation algorithm. The designed switch is fully parallel, uses a neural network for switch management and also for traffic engineering. The switch uses an advanced packet dropping mechanism. The article describes the switch design (network processor design) and compares the developed switch with other “conventional” architectures. The subject of comparison is the architectures and performance.

Brief Biography of the Speaker:
Vladislav Skorpil, PhD. attended the Brno University of Technology, Faculty of Electrical Engineering, Department of Telecommunications. He graduated from this university in 1980. From 1985 to 1989 he was a doctoral student in the same Department. From 1980 to 1982 he worked as a designer for the telecommunication design office. He again entered the Department of Telecommunications of BUT in 1982 as a university teacher and he has been working in this department since that time, now he has been a vice-head of this department and Assoc.Prof. (from 2004). He takes a keen interest in modern telecommunication systems. He has taught in courses on transmission systems from analogue through all categories of digital up to special applications. He is the author of about 105 international scientific papers and some manuals. He has complemented his theoretical knowledge by co-operation with a lot of firms and institutions. He has co-operated on telecommunication projects such as digital transmission and switching systems, telecommunication broadband networks, ISDN, ATM, data networks LAN and MAN, on structured cabling design, neural networks, wavelet transform, Quality of Service QoS, data bit rate compression, etc. He is a member of international organisations IEEE and WSEAS.