Machine learning techniques such as deep neural networks have
become an indispensable tool on an increasing number of complex tasks due to the
availability of large databases and recent improvements in deep learning
methodology. However, these models are usually applied in a black box manner because
of their nested non-linear structure, which limits its applications in numerous
scenarios. Therefore, it is imperative to develop methods for explaining and interpreting
the current machine learning models. Currently, we intend to carry out a systematic investigation of
how to build an interpretable artificial intelligence framework by bridging the
gap between the abstract latent features and human concepts, with a set of
sound theoretical analysis and extensions in various application.
