Rectangular VLSI cell placement using force directed self-organizing maps and delta learning rules

VLSI cell placement, with various constraints on cell size, wire connections and layout dimensions, is complicated and has been proven to be NP-hard. In a previous paper (IEEE Trans. Neural Networks, 1993), we described a force-directed self-organizing map (FDSOM), based on Kohonen's (1989) algorithm, for resolving the VLSI cell placement problem using the circle cell model and the perceptron learning rule. In this paper, a delta learning FDSOM is presented to resolve the cell placement problem with the rectangular cell model. Experiments show that the total wire length obtained is shorter than that obtained by previous methods.<
>