Classifiers applied to dimensioning of splitters in PON design

We present an approach to designing passive optical access networks (OANs) that optimizes the selection of splitting combinations in the distribution network. The proposed methodology uses classifiers (a form of pattern matching) to select splitting combinations for particular network clients. The methodology optimizes the classifiers instead of directly optimizing splitting combinations for each client independently. We numerically evaluate the approach and show conditions under which it reaches its full potential. We show that this design method can be combined with direct optimization heuristics to increase its efficiency. In addition to demonstrating its usefulness in creating better passive optical network designs, we show how the obtained solutions from one design can be reused when solving similar passive OAN optimization problems in the future.