Abstract: |
Additional test cases are added one by one, as the population solves the current set of test cases, until some fixed final limit is reached. We examine the general nature of this approach. Complexity is defined in a general sense. It is proved that adding a test case to a test set never reduces the complexity of a solution, and never increases the probability of finding a solution. The terms representative and redundant, are formally defined. The variation in the number of test cases and the jumps in the number of test cases are observed. The size of the test set just before a general solution is found, indicates a threshold number of test cases required for generalization. We observe, how generalization varies with the size of the test set. Finally we observe the number of successes per evaluation required to produce a general solution. |