The Microsoft New Faculty Fellowship Program was established to recognize junior faculty who are advancing computing research in novel directions, have the potential for high impact on the state of the art, and demonstrate the likelihood of becoming innovative leaders in the field. The selection process was carried out across the United States and Canada in many computing areas. It was a is highly competitive multiple-step process: Every institution nominated one candidate with strong internal and external recommendation letters. Among approximately 100 nominees from different institutions and research areas, 11 candidates were invited for on-site interviews. Through this process, Assistant Professor Yanlei Diao was selected as a finalist for the Microsoft New Faculty Fellowship with an award trophy to recognize her research accomplishments and vision.
Diao’s research is based on the vision that recent advances in sensing, network monitoring, and application monitoring will soon generate an unprecedented volume of real-time streaming data. Hence, there is an increasing demand of an information infrastructure that collects real-time data and delivers meaningful, actionable information of vital economic, social, and environmental importance. A large class of emerging applications in this new infrastructure presents daunting yet critical challenges that call for significant broadening and enriching of data management research. For example, in many applications such as object tracking and severe weather monitoring, data is incomplete, noisy, and even erroneous, and such data is produced at high rates. As such data is passed through various processing stages, the results are often of unknown quality. There is an increasing need for applications and infrastructures that can deal with such uncertainty. Similarly, there is a growing need for applications that can efficiently manipulate and handle low-quality data. An increasingly important form of data-information transformation is to define application information needs using complex logic that uses filtering, correlation, and sophisticated pattern matching. Such transformations must be performed in real-time and often using low-quality data.
To address these challenges, Diao proposes two new directions of research in data management: To handle uncertain data, her research advocates a fundamental extension of real-time data management systems with learning capabilities, including efficient inference over data streams, modeling computation error, and self-tuning for error reduction. For complex data-information transformation, her research explores Complex Event Processing, a new processing paradigm that is grounded in automata and complexity theory and amenable to efficient, robust implementation.
By way of addressing these issues, her research has the potential to bridge the gap between the requirements of a new class of applications and the state-of-the-art with theoretical, algorithmic, and systems contributions.