My research focuses on visual analytics solutions to understand deep learning models. Deep learning has achieved remarkable results at many intelligent tasks in areas of image, speech and natural language. This means that these models have found a solution for a certain task. However, the question what the solution is remains unanswered. The challenge of anwswering this question comes partly from the fact that we don’t know what we are looking for. Oftentimes we don’t even have a clear picture of how human solves the same problems, such as recognizing an image or understanding language, which makes it hard to formulate or validate theories of how a neural network solves these problems. Answers to this question can faciliate training, debugging exisiting models, creating better models and inspiring ideas of how human brain works. Visual analytics marries the powers of automatic data analysis and human visual perception to support exploration and dicovery of large datasets and complex processes. I believe that effective and creative design of visual analytical systems can enable us to see the patterns hiding in deep neural networks.