We survey the application of a relatively new branch of statisticalphysics--"community detection"-- to data mining. In particular, we focus on thediagnosis of materials and automated image segmentation. Community detectiondescribes the quest of partitioning a complex system involving many elementsinto optimally decoupled subsets or communities of such elements. We review amultiresolution variant which is used to ascertain structures at differentspatial and temporal scales. Significant patterns are obtained by examining thecorrelations between different independent solvers. Similar to othercombinatorial optimization problems in the NP complexity class, communitydetection exhibits several phases. Typically, illuminating orders are revealedby choosing parameters that lead to extremal information theory correlations.
展开▼