No label defined (Q3827451)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 4102352
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 4102352 |
Statements
1987
0 references
pattern inference
0 references
Bayesian approach
0 references
method of sieves
0 references
penalized likelihood
0 references
0.7336413264274597
0 references
0.7174032926559448
0 references
0.7149669528007507
0 references
0.7126272916793823
0 references
[For the entire collection see Zbl 0671.00013.] The literature on pattern research, which is at present expanding at an accelerating rate, contains numerous algorithms intended for the recognition of patterns of different types, among them especially pictorial patterns. Common to most of these algorithms is that their construction is based on plausibility arguments and only seldom rests on a firm mathematical foundation. Nevertheless, some of them seem to work well when tried on real or artifically generated patterns. Actually, it is difficult to evaluate their performance in some exact terms just because of their lack of precisely stated assumptions and mathematical justification. We have argued for a long time that the algorithms intended for pattern inference (restoration, recognition, extrapolation, etc.) should be derived from precisely formulated models of pattern ensembles: the algorithms should be derived as solutions, exact or approximate, to optimality problems formulated in terms of an explicitly stated criterion. Viewed from a general statistical perspective it turns out that the problems encountered with our approach are of a type dealt with in abstract inference, since both parameter and sample space will turn out to be of high or infinite dimensionality. In this paper we shall use exclusively a Bayesian approach, but others are sometimes indicated, for example the method of sieves and penalized likelihood.
0 references