Quality Evaluation Assurance Levels for Deep Neural Networks Software

Quality of machine learning software products or services is dependent on datasets used for training. However, defining quality of datasets is difficult, which might bring about risks in business situations. Independent, third-party testing laboratories would mitigate the risks. This paper proposes quality evaluation assurance levels, which is a basis of a third-party evaluation and certification framework. Moreover, quality of machine learning software is indeed viewed from three perspectives: prediction performance quality, training mechanism quality, and lifecycle support quality enabling continuous operations.