Machine Learning

Xinyun Chen is a third-year Ph.D. student at UC Berkeley, working with Prof. Dawn Song. Her research lies at the intersection of deep learning, programming languages, and security. Her recent research focuses on neural program synthesis and adversarial machine learning, towards tackling the grand challenges of increasing the accessibility of programming to general users, and enhancing the security and trustworthiness of machine learning models. She received her bachelor’s degree from Shanghai Jiao Tong University, and has interned at Facebook AI Research and Google Brain.

For neural program synthesis, she develops deep learning techniques to better leverage syntactic and semantic information of programs, so as to synthesize programs with better generalizability and higher complexity, and from different types of specifications. She has been working on: (1) translating natural language descriptions to programs; (2) program synthesis from input-output examples; and (3) leveraging deep learning techniques for software engineering applications.

For adversarial machine learning, her work mainly focuses on exploring the vulnerabilities of existing machine learning models, to enable a better understanding of the model behavior, and shed some light on defense proposals to enhance the robustness of the state-of-the-art models. In particular, she has worked on black-box attacks in various scenarios, where she demonstrates that an adversary with weak capabilities may still be able to successfully launch the attacks, suggesting that such attacks could pose threats in the real world.