One-Size-Fits-None? Improving Test Generation Using Context-Optimized Fitness Functions

摘要

Current approaches to search-based test case generation have yielded limited results in terms of human-competitiveness. However, effective search-based test generation relies on the selection of the correct fitness functions—feedback mechanisms—for a chosen goal. We propose that the key to overcoming these limitations lies in infusing domain knowledge and context into the fitness functions used to guide the search and the ability to automatically optimize the fitness functions used when generating tests for a given class, goal, and algorithm.