Towards dependable and explainable machine learning using automated reasoning

The ability to learn from past experience and improve in the future, as well as the ability to reason about the context of problems and extrapolate information from what is known, are two important aspects of Artificial Intelligence. In this paper, we introduce a novel automated reasoning based approach that can extract valuable insights from classification and prediction models obtained via machine learning. A major benefit of the proposed approach is that the user can understand the reason behind the decision-making of machine learning models. This is often as important as good performance. Our technique can also be used to reinforce user-specified requirements in the model as well as to improve the classification and prediction.No Full Tex