Collisions are helpful for computing unique input-output sequences

Computing unique input-output sequences (UIOs) from finite state machines (FSMs) is important for conformance testing in software engineering, where evolutionary algorithms (EAs) have been found helpful. Previously, by using a fitness function called W-fitness, (1+1)-EA was theoretically shown to be superior to random search on some FSM instances. Motivated by the observation that many plateaus exist in the fitness landscape of the W-fitness function, in this paper, we propose a new fitness function called C-fitness which is able to override the plateaus through exploiting collisions among the states of FSMs. We theoretically analyze the running time of (1+1)-EA on two problem classes. Our results show that the performance of (1+1)-EA using C-fitness is generally better and never worse than that using W-fitness in our studied cases, implying the importance of exploiting problem structures. Categories and Subject Descriptors F.2 [Theory of Computation]: Analysis of Algorithms ...

• Type: Journal
• Year: 2011
• Where: GECCO
• Authors: Chao Qian, Yang Yu, Zhi-Hua Zhou
• Keywords: Computation Analysis, Conformance Testing, GECCO 2011, Optimization, Problem Complexity