Copyright 9/29/1999 ............. ................................... ................................... ............. http://www.cs.ucla.edu/~klinger/pami/summ.html
1. A pattern is an element from a set of objects that can in some useful way be treated alike.
2. A feature is a pattern-attribute useful for classification decision-making.
3. Pattern instances differ in details but are similar in general appearance or underlying nature.
4. Pattern analysis/recognition tasks involve sets.
5. One pattern analysis/recognition approach uses probability and statistical ideas.
6. Another focuses on structural issues.
7. Pattern recognition is one of five central subjects in artificial intelligence.
8. Learning is another of the five, which include search, planning and induction.
9. A computationally costly decision-rule for classifying an unknown-class pattern is nearest-neighbor.
10. A similar less-costly decision-rule uses prototypes (templates).
11. Pattern prototypes could be just the mean (simple average) of their numerical representations.
12. Calculating sample-mean (variance) of multidimensional vectors is done analogously to scalars.
13. Probability involves a universe of possible occurrences and a sample space listing them all.
14. For dealing with more than one random characteristic at a time use and (comma in symbols).
15. For restriction to part of the occurrences universe use given that (vertical line in symbols).
16. Expected value is the mathematical probability-weighted sum or integral of a random quantity.
17. A vector sample mean m(n) of n patterns x1, x2, ..., xn sums them and divides by n.
18. Histograms and scatter-diagrams can picture the general or set properties of patterned data.
19. Patterns with class labels are the basic given data in constructing a decision-rule.
© 1999, Klinger, Allen