Neural network architectures are effectively applied to solve the channel routing problem. Algorithms for both two-layer and multilayer channel-width minimization, and constrained via minimization are proposed and implemented. Experimental results show that the proposed channel-width minimization algorithms are much superior in all respects compared to existing algorithms. The optimal two-layer solutions to most of the benchmark problems, not previously obtained, are obtained for the first time, including an optimal solution to the famous Deutch's difficult problem. The optimal solution in four-layers for one of the be lchmark problems, not previously obtained, is obtained for the first time. Both convergence rate and the speed with which the simulations are executed are outstanding. A neural network solution to the constrained via minimization problem is also presented. In addition, a fast and simple linear-time algorithm is presented, possibly for the first time, for coloring of vertices of an interval graph, provided the line intervals are given.
Library of Congress Subject Headings
Telecommunication--Switching systems--Data processing; Neural networks (Computer science)--Programming; Algorithms
Department, Program, or Center
Computer Engineering (KGCOE)
Islam, Taj-ul, "Channel routing: Efficient solutions using neural networks" (1993). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus