Tian Guo, assistant professor of computer science, has received a three-year, $499,723 award from the Sponsor: National Science Foundation for a project titled, “CSR: Small: Towards Efficient Deep Inference for Mobile Applications.” Guo will research ways to support deep learning-based mobile applications and then use that information in teaching her own courses. This is an integrated project that includes research, outreach, and education.
Deep learning, which is in the machine learning family, is used in computer vision, speech recognition, natural language processing, and gaming. It also allows mobile applications to provide novel and useful features, making the apps “smarter.” However, the current way to use deep learning models to perform inference tasks in mobile applications leads to poor user experience.
Guo’s project is aimed at making it easier for mobile app developers to do their job and to provide a better experience for app users. This project should develop and demonstrate the benefits of flexible, fine-grained mobile deep inference with pre-partitioned models and dynamic inference execution plans. It also should design and train a collection of mobile-specific deep learning models, as well as provide a cloud-based deep learning model layer manager. The program implements an educational plan that exposes undergraduates and K-12 students to research in data science. Two WPI PhD students are working on this project and undergrads are expected to participate, as well.