Adaptive weighting method for layout optimization with multiple priorities
AssigneeInternational Business Machines Corporation
InventorsMichael S. Gray, Matthew T. Guzowski, Kevin W. McCullen, Xiaoping Tang, Robert F. Walker, Xin Yuan
Life of this patent
Estimating the term…
Abstract
An adaptive weighting method for layout optimization differentiates different priorities by assigning the weight of a higher priority (pi) to be multiple of the weight of a lower priority (pi−1) where W(pi)=mi % W(pi−1. To avoid numerical imprecision, this method keeps the total cost in the objective function within a trustable range by scaling the initial weights in the objectives, while maintaining relativity, to produce the scaled weights.
Claims6
A method of adaptive weighting for layout optimization of multiple objectives having priorities, comprising:
- setting initial minimum perturbation weights;
- receiving, as input, priorities and trade offs between said priorities;
- calculating objective weights based on said priorities and said trade offs;
- modifying said objective weights to bias objectives having a same priority;
- calculating an initial total objective cost based on said priorities, said trade offs and said minimum perturbation weights and said objective weights;
- scaling said initial minimum perturbation weights and said objective weights to maintain subsequent total objective costs in a numeric range, while maintaining relativity, to produce scaled perturbation weights and scaled objective weights;
- solving said layout optimization using said scaled perturbation weights and said scaled objective weights to produce a result;
- and
- outputting said result.
A method of adaptive weighting for layout optimization of multiple objectives having priorities, comprising:
- setting initial minimum perturbation weights;
- receiving, as input, priorities and trade offs between said priorities;
- calculating objective weights based on said priorities and said trade offs;
- modifying said objective weights to bias objectives having a same priority;
- calculating an initial total objective cost based on said priorities, said trade offs and said minimum perturbation weights and said objective weights;
- scaling said initial minimum perturbation weights and said objective weights to maintain subsequent total objective costs in a numeric range, while maintaining relativity, to produce scaled perturbation weights and scaled objective weights, wherein said scaling comprises assigning a weight of a higher priority to be a multiple of a weight of a lower priority;
- solving said layout optimization using said scaled perturbation weights and said scaled objective weights to produce a result;
- and
- outputting said result.
Dependencies are inferred from the claim text.