Pattern synthesis
Ulf Grenander · 1994
Abstract Having introduced a regularity controlled probability measure on a configuration space the question naturally arises: can one simulate it? This task, that we shall speak of as pattern synthesis, is important for two reasons. First, during the model-building stage it is imperative to get some intuitive idea of the sort of patterns that are (typically) generated by the measure and to see how their behaviour varies as we change the structure of the model. It is difficult to predict the pattern behaviour directly from the construction of generator space, bond relation, similarity group, etc. To learn about this we shall simulate the measure and display the resulting patterns graphically or otherwise.