Solving Sequencing Problems Using Recurrent Neural Networks and Simulated Annealing - a Structural and Computational Comparison -

Since most sequencing problems are NP-hard, many heuristic techniques have been tried to solve such problems. In this paper, we investigate two methods Hopfield neural networks and Simulated Annealing and their application to the problem of placing standard cells during VLSI design. Comparisons were made with respect to solution quality and computing time. Simulated Annealing performed better in all test cases, but the differences between solution qualities decrease as problem sizes and complexities are higher. Advantages and disadvantages of both methods, and approaches to improve their performance are discussed.