Abstract

This paper discusses a theoretic-informational approach to the problem of pattern recognition. Using methods of generalized decision functions, support vectors, and finite mixtures as an example, it is shown that the existing quality criteria of the decision rules can be represented as particular implementations of the principle of minimum description length. This makes it possible to refine the available criteria and thereby improves the discriminant properties of the indicated recognition methods and allows recognition systems to be constructed with automatic selection between particular recognition methods based on a general quality criterion. The indicated ideas are successfully verified experimentally, using as an example the problem of recognizing small targets.

Modeling the temporal evolution of an aero-optical aberration with the minimum description length principle

Qiong Gao, Zongfu Jiang, and Shihe Yi
Opt. Lett. 39(11) 3126-3129 (2014)

Minimum Description Length approach for unsupervised spectral unmixing of multiple interfering gas species

Julien Fade, Sidonie Lefebvre, and Nicolas Cézard
Opt. Express 19(15) 13862-13872 (2011)

Improved method for estimating the minimum length of modal filters fabricated for stellar interferometry

Sonali Dasgupta, N.G.R. Broderick, David J. Richardson, Tomer Lewi, and Abraham Katzir
Opt. Express 17(3) 1935-1946 (2009)

Discriminant analysis for recognition of human face images

Kamran Etemad and Rama Chellappa
J. Opt. Soc. Am. A 14(8) 1724-1733 (1997)

Multi-class remote sensing object recognition based on discriminative sparse representation

Xin Wang, Siqiu Shen, Chen Ning, Fengchen Huang, and Hongmin Gao
Appl. Opt. 55(6) 1381-1394 (2016)