There are two basic approaches to pattern recognition: decision-theoretic and syntactic. However, in actual applications, a combination of both may be needed. One such hybrid technique consists of syntactic method coupled with stochasticity in its grammar. Randomness in the syntactic case is caused due to noise and insufficient information about characteristics of pattern classes. To absorb the effect of this randomness, the grammar must be generalized to include the probabilities of production rules.
In this paper, a preliminary discussion of issues involved with hybrid techniques, in general, and stochastic grammars, in particular, is provided. An efficient algorithm for an automatic learning of production probabilities is devised. Concepts are illustrated via examples.
| Identifer | oai:union.ndltd.org:auctr.edu/oai:digitalcommons.auctr.edu:dissertations-4098 |
| Date | 01 May 1987 |
| Creators | Placide, Eustache |
| Publisher | DigitalCommons@Robert W. Woodruff Library, Atlanta University Center |
| Source Sets | Atlanta University Center |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | ETD Collection for AUC Robert W. Woodruff Library |
Page generated in 0.0017 seconds