Weakened Generalized Precedents in Problems of Searching for Informative Aspects of Data

Generalized Precedents (GPs) employ practical expertise when it's expressed as a pattern in data in terms of few parameters. It's a multidimensional analog of the Hough transform, so the targets are sharp peaks of the distribution in parametric space. Weakened GPs use less exact description of the sought pattern, and transfer the analysis onto the preferred shape of the secondary clusters themselves (not only peaks). It's shown that useful information can be mined this way exceptionally on the base of some natural a priori expertise of the mentioned kind (smoothness, curvature, uniformity).