2018/01/03 by Peter Moeck, Moeck, Peter
Computer Science · Materials Science · #Applied Physics (physics.app-ph) #Computational Drug Discovery Methods #FOS: Physical sciences #Machine Learning in Materials Science #Materials Science (cond-mat.mtrl-sci) #X-ray Diffraction in Crystallography
paper · pdf · doi:10.48550/arxiv.1801.01202
openalex publication_date 2018/01/03 · openalex created_date 2025/10/24 · openalex updated_date 2026/07/28
Geometric Akaike Information Criteria (G-AICs) for generalized noise-level\ndependent crystallographic symmetry classifications of two-dimensional (2D)\nimages that are more or less periodic in either two or one dimensions as well\nas Akaike weights for multi-model inferences and predictions are reviewed. Such\nnovel classifications do not refer to a single crystallographic symmetry class\nexclusively in a qualitative and definitive way. Instead, they are\nquantitative, spread over a range of crystallographic symmetry classes, and\nprovide opportunities for inferences from all classes (within the range)\nsimultaneously. The novel classifications are based on information theory and\ndepend only on information that has been extracted from the images themselves\nby means of maximal likelihood approaches so that these classifications are\nobjective. This is in stark contrast to the common practice whereby arbitrarily\nset thresholds are employed to force crystallographic symmetry classifications\ninto apparently definitive/exclusive states, while the geometric feature\nextraction results on which they depend are never definitive in the presence of\ngeneralized noise, i.e. in all real world applications. Thus, there is\nunnecessary subjectivity in the currently practiced ways of making\ncrystallographic symmetry classifications, which can be overcome by the\napproach outlined in this review.\n