On modeling

Everyone is used to certain types of models, like probabilistic models, GM, ODE and e.t.c. What is rarely spoken about is a particular class of models - computational once. Usually, people refer to computational model when they speak of obtaining a numerical solution for a good old model based on DE or whatever, but that is just dull. What is really exciting - a model defined by computation from the very beginning. Some segmentation algorithms clearly define segments by means of computational modeling - segment is a result of algorithm provided. What made me think about it is a presentation by Olga Sorkine from NYU on "iWIRES: An Analyze-and-Edit Approach to Shape Manipulation". There, the manifold of plausible shapes is defined by the convergence of edit operations applied to shape initialization.