<H2><B>Exercises 2</B></H2>
10/18/99 Version www.cs.ucla.edu/~klinger/pami/exer3.html
CS 276A Pattern Analysis and Machine Intelligence
A. Klinger Fall 1999
Exercises 3

0. Design a six-by-six patterned array on four symbols. (The image should have some regular structure, yet not be entirely fixed by that.) .

1. Give an example where different costs (loss) are associated with the possible class-labeling (recognition) decisions. (Avoid the medical and seismic areas discussed in class.

2. In a random experiment with some chance outcomes the technical term for average is expected value. Explain why the numerical result we call "expected value" could be something that would never occur in practice.

3. Explain the difference between a minimum risk and a Bayes Rule decision procedure in words.

4. Refer to http://www.cs.ucla.edu/~klinger/pami/six.html . Complete part a) there (complexity).

5. Refer to http://www.cs.ucla.edu/~klinger/pami/six.html . Complete part b) there (order/harmony).

6. Refer to http://www.cs.ucla.edu/~klinger/pami/six.html . After completing 4. and 5., combine them. I.e., complete part c) at the above locator (aesthetic measure).

7. Create features that would differentiate between patterns similar to those in http://www.cs.ucla.edu/~klinger/pami/six.html . In other words, assume that each of the six are prototypes or templates. Your features should be useful for making class assignment decisions when presented with other patterns from each of the six classes.