A Short Tutorial on Reconstructability Analysis
Systems Science Ph.D. Program, Portland State University
Last modified: May 18, 2006
This talk will present the basic ideas of Reconstructability Analysis. RA is a general modeling methodology developed in the systems community since the early 60s (pioneered by Ross Ashby) for exploratory (data mining, machine learning) and confirmatory modeling. RA is based in graph theory, set theory, information theory, and statistics. It is applicable to qualitative (nominal) and, via binning, to quantitative (continuous) data, with both statistical & non-statistical applications. It overlaps and augments more widely known methods, e.g., log-linear modeling and Bayesian networks.