On the Computational Complexity of Parameter Estimation in Adaptive Testing Strategies

Abstract

Adaptive testing is the counterpart of adaptive control in software testing. It means that software testing strategy should be adjusted on-line by using the testing data collected during software testing as our understanding of the software under test improves. In doing so, online estimation of parameters plays a crucial role. In this paper, we investigate the computational complexity of the parameter estimation process in two adaptive testing strategies which adopt different parameter estimation methods, namely the genetic algorithm (GA) method and the recursive least square estimation (RLSE) method. Theoretical analysis and simulations are conducted to compare the asymptotic complexity and the runtime overhead of the two adaptive testing strategies. Finally, a controlled experiment on the Space program is conducted to measure the relationship between computational complexity and the failure detection efficiency for the two strategies.