674 CHAPTER 11 Placement where area(e) is the total area of all modules in e. In addition, modules are vis- ited in ascending order of module area (with preference to smaller module degree to break ties). This ordering is observed to balance the area of clusters better. If a good initial placement is provided, the proximity information between modules can be incorporated into the affinity as follows: r ij ¼ c e ð j e j À 1 Þ Â area ð e Þ Â dist ð i; j Þ e2E^i; j2e X where dist(i, j ) is the Euclidean distance between i and j. 11.5.4.2 Best choice In the Best Choice clustering technique [Alpert 2005], the affinity is defined as: r ij ¼ c e j e j  ð area ð i Þ Â area ð j Þ Þ e2E^i; j2e X where area(i ) and area( j ) are the areas of modules i and j, respectively. In addi- tion to the indirect control of the cluster size by the affinity, Best Choice imposes a hard upper limit for cluster size. Moreover, the pair of modules with the largest