About Me
Dr. Mengdi Huai is currently an assistant professor in the Department of Computer Science at Iowa State University. Her research interests are in the areas of data mining, machine learning, and artificial intelligence (AI). Currently, her work focuses on Responsible AI, where she investigates how to identify risks and develop mitigation strategies throughout the data mining and machine learning lifecycles. This includes exploring key aspects of trustworthiness, such as model interpretability, uncertainty, security and robustness, fairness, and privacy. Additionally, she is also interested in designing effective data mining and machine learning algorithms to deal with complex data with both strong empirical performance and theoretical guarantees.
Why I Like My Research
I hope I can use my knowledge to solve practical problems and change people’s lives for the better. My current research mainly focuses on developing novel techniques to make machine learning more trustworthy. Machine learning has been applied widely to build intelligent systems, such as self-driving cars, intelligent recommendation systems and clinical decision support systems. However, in the real world, these systems suffer from a trust crisis due to the lack of trustworthiness behind their behaviors. In the future, I hope I can address this challenge and help people benefit from machine learning-based intelligent systems.
Success in My Group
Success in my group is driven by curiosity, hard work, collaboration, open communication, and shared goals. Each student in my group has unique strengths, creating a diverse skill set that enhances our overall effectiveness. We celebrate each achievement, big or small, fostering an environment where everyone feels valued and motivated. This collaboration not only propels us toward our goals but also deepens our connections, making our group a place where personal growth and collective success thrive together.
Example Research Projects
Some of the research projects REU students can work on with me include:
Model Transparency for Predictive Models
Users usually treat a machine model as a black box due to its incomprehensible functions and complex working mechanism. The “black box” nature of machine learning models may impede decision makers from trusting the predicted results, especially when the model is used for making critical decisions (e.g., medical disease diagnosis), because the consequences may be catastrophic if the predictions are acted upon blind faith. In this project, we will explore predictive explanations to faithfully reflect the model behavior during the decision-making process, which helps users examine whether a machine learning model has employed the true evidences instead of biases and reduces the likelihood of an error.