This repository contains the poster and report that submitted to the project for graduation at Fall semester, 2019 held in Dept. of Computer Science and Engineering, Korea University.
Neural networks have become more important than ever. However, testing techniques did not evolves compare to the that of the neural networks. We showed that current testing techniques have some limitation because of their selection strategies, and suggest new technique, called Adapt, that adaptively generates selection strategy with respect to network to test, input seeds for testing, and coverage metrics for evaluation.