T Cell–Inspired Algorithms - Immunological Computation: Theory and Applications - page 111

4.11 Research Questions
1. Study the characteristics for matching rules.
2. How hard is it to fi nd detectors that do not match self with diff erent
representations and matching rule?
3. How many detectors are required providing reasonable protection?
4. How does the number of detectors scale with size of self?
5. How to pick l , r , N R , and other parameters in binary NS algorithm?
6. Develop a suitable test suite for evaluating NS algorithms?
7. How to address various applicability issues of real-valued NS algorithms?
8. Develop a real-world application of NDB. Compare the results with other
methods.
9. Design and implement a new NDB algorithm.
10. Use various negative operations (algebra) to implement an e cient NDB.
4.12 Review Questions
1. What are diff erent NSAs? Which is the main goal of these algorithms?
2. What are the main features of the NS process that occurs in biological
immune systems?
3. Which immune natural process is simulated by artifi cial NS?
4. Write a pseudocode showing the major steps of an NSA.
5. List some real-world applications where NSAs could be used.
6. Why only samples from one class are su cient for an NSA?
7. What are the main stages in an NSA?
8. Classify the following statements into either describing detector generation
phase or monitoring/testing/detection phase:
A set of detectors are generated by some randomized process.
Candidate detectors that match any of the self-samples are eliminated,
whereas a subset of the remaining ones is kept.
Matching rules are based on T cell/antigen a nity measures.
Detectors are used to check whether new incoming patterns correspond to
self or nonself instances.
If an input pattern matches a detector, then it is identifi ed as part of
nonself.
9. Give a specifi c problem and representation where a binary representation is a
better choice than a real-valued vector representation.
10. What are the advantages of binary representation over real-valued vector rep-
resentation, if any?
11. What are the advantages of real-valued vector representation over binary rep-
resentation, if any?