In the big data era, the rapid rise of deep learning has made great progress in many fields, such as computer vision. In recent years, with the accumulation of software artifacts, deep learning has begun to play an important role in software engineering domain. We review recent advances of deep learning methods in software analysis, summarize major research interests and application features. A batch of important results has been published, and related research shows an upward trend. In the end, limitations and problems in the application of deep learning methods are discussed.