The distributed stochastic method of large-scale optimization with organization

本计画研究发展一种新的以生物现象为基础的随机最佳化方法，并应用于超大规模积体电路的布局问题，实际上，我们将模拟退火方法与生物进化的变异与交叉概念结合。由于这种组合使得在成本景观中自由移动成为可能，因此我们可以尝试在大区域中搜索全局最小值。这种方法使并行计算成为可能。本文从近似的角度出发，研究了平均场理论，得到了一些加快随机优化速度的有用结果，并将本文提出的方法应用于VLSI布图规划、布线和布局等问题中。仿真结果验证了该方法的有效性和实用性。

The project has studied the development of a new stochastic optimization method based on biological phenomena and applied to VLSI layout problems.Practically speaking, we combined the simulated annealing method with mutation and crossover concepts of biological evolution. Since this combination makes it possible to move freely in the cost landscape, we can try to search a global minimum in the large area. This method makes parallel computation possible. If we could implement on parallel processors, the computation time of the proposed method would be very fast.From the viewpoint of approximation, we studied mean field theory and got some useful results on the speed up of stochastic optimization.We applied the proposed method to the layout of VLSI, such as floorplanning, routing and placement problems. The simulation verified the effectiveness and usefulness of the method.