Lower Bounds on the Size of Test Data Sets.
Tim Menzies, Sam Waugh · 1998
Practically speaking, how small can a test suite be and still be of value? In the context of temporal graph-theoretic abductive validation, the answer to this question is very language-dependent. Seemingly trivial variations in a language can have a significant impact on how large a test suite must be. This paper is hence a cautionary note to those who invent languages and ontologies without experimentally testing the practicality of those languages. Submitted to the Validation and Verification of Knowledge-Based Systems, 1 June 1998, in conjunction with the Sixth International Conference on Principles of Knowledge Representation and Reasoning (KR'98) Trento, Italy, June 2-5, 1998. 1 Introduction Validating a theory is hard work. Validation is complicated even further in poorly-measured domains. In such domains, the cost of data collection prevents us collecting all the observations we desire. Such observations could be used to validate that a theory of X can reproduce know...