Research summary
Machine learning for prediction systems is powered by human-annotated data, which add subjectivity, intent and clarification. But these data can also become a bottleneck-and sharing information can create significant ethical and security concerns. Fortunately, the emergence of digital data is creating new opportunities both to break the bottleneck and to address these concerns. Dr. Yuhong Guo, Canada Research Chair in Machine Learning, is figuring out how.
She and her research team are developing more effective methods of sharing information across diverse and changing real-world scenarios. This will reduce the dependence on human annotation for supervised learning while also addressing privacy and fairness considerations. Ultimately, their work will support a new generation of information transfer technologies that allow machine learning to be applied in diverse, adaptable and socially beneficial services.