What are the samples for learning efficient routing heuristics? [MCM routing]

In this paper we present a genetic algorithm (GA) based approach to learn heuristics for MCM routing. More exactly, given the MCM, the GA learns heuristics for routing this specific MCM. These heuristics are sequences of some given basic optimization modules (BOMs). The learning environment consists of a set of routing samples, which we call the training set. Since the training set plays a key role in the learning process, the paper is focused on the search for proper training sets. Two methods which are based on hierarchical decomposition of MCMs are proposed and experimentally proven to be very efficient. The experiments show that efficient and fast heuristics for large problems can be rapidly learned by GAs starting with problem specific BOMs, in general.