Selection of test samples to improve DNN test efficiency based on neuron clusters

Deep neural networks (DNNs) have complex structures and operation methods, making it difficult to pinpoint the cause of performance degradation. Moreover, DNNs are widely used in various domains, and the significance of their quality is emphasized. In the test process, selecting test samples that cause model misclassification can detect model vulnerabilities at an early stage. Improving test efficiency, defined as the number of misclassified samples compared to the number of selected test samples, can reduce the time and cost required for testing. Therefore, it helps to perform more tests within a limited time and cost. In existing sample selection studies to improve test efficiency, test samples are selected based on coverage. However, samples achieving high coverage do not necessarily guarantee improved test efficiency. This paper proposes a test sample selection method for improving test efficiency based on neuron clusters. The neuron cluster is a simple and effective concept that abstracts DNN’s computational process for human understanding. The most significant contribution of this paper is to improve test efficiency by selecting samples that can be misclassified based on neuron clusters. Through experiments using public datasets and ResNet models, we compare the proposed method with coverage-based test sample selection methods in terms of the test efficiency. Our proposed method is at least 8%p more efficient than the coverage-based test sample selection methods, although it varies somewhat among datasets. We can apply neuron clusters in various ways to DNN testing, including identifying untrained class samples and selecting test samples.