Deep learning (DL) is fundamentally intertwined with software engineering (SE). The application of DL to a computational problem represents a new programming paradigm: rather than developing a program in code, a program is “learned” from large-scale datasets. This can be referred to as SE for DL. On the other hand, DL techniques can be used to automate or improve existing SE tasks. This is referred to as DL for SE. In this case, DL systems can be viewed as an inflection point for software development, as they enable new capabilities that cannot be realized cos-effectively through “traditional” software wherein the behavior of a program must be specified analytically. There currently exists an unprecedented amount of software data that is freely available in ope-source software repositories. This data spans several software artifacts, from source code and test code, to requirements and issue tracker data. Given the effectiveness by which DL systems are able to learn representations from such largescale data corpora, there is ample opportunity to leverage DL techniques to help automate or improve a wide range of developer tasks. All these new forms of development carry with a new set of challenges that represent several opportunities for novel research.

The topics of interest in the discussion are but not limited to:

• Intersection of DL and SE
• DL techniques used in SE
• DL-based models for software
• NLP, CV, and multimodal technologies for SE
• DL-based recommendation systems in SE
• DL components in software or systems
• Practicability in utilizing DL techniques for SE
• Effectiveness, efficiency, understandability, and testability of DL-based solutions
• DL techniques used to boost performance for various SE tasks
• The impact of DL model selection, optimization, and structure differences on SE
• Employment of neural networks on SE projects
• Best practices for building DL systems
• Software processes supported by DL models
• A set of current challenges remaining to be investigated