Abstract: The complex system is a system of systems, containing a group of interacting entities that form a unified whole. The main missions of AI for complex systems are understanding the systems’ characteristics, diagnosing systems’ problems, and controlling systems’ behaviours. Although many AI technologies are being developed for complex systems, how to make them trustable to fulfil their missions remains a problem. In this talk, I will explore ways to achieve trustable AI for complex systems, where two typical complex systems, life systems and IT systems, are discussed. Ways to achieve trustable AI are summarized into three categories, which are making data trustable, making a good understanding of complex systems and making AI algorithms trustable. I will first talk about how to make data trustable in three aspects, which are seeking deeper, wider and bigger data. The approach for automated phenotype annotation on electronic health record will be introduced in particular. Then, I will introduce the holistic view of the complex systems for a good understanding. Finally, I will discuss approaches to make AI algorithms trustable, which lead to my future research directions.
Bio: Dr. Xian Yang is a Research Fellow at the Data Science Institute at Imperial College London. She had been working as a research associate/assistant at the Data Science Institute from 2012 to 2018. During that period, she has taken part in many cross European research projects, such as UBIOPRED (severe asthma subtyping), eTRIKS (knowledge management platform for translational medicine), Optimise (multiple sclerosis disease prognosis and treatment) and iHealth (clinical treatment pathway optimisation). Her main role in these projects was developing data analysis/machine learning methods to analyse and construct predictive models from OMICS, clinical and other datasets such as survey data. From 2018 to 2019, she worked in Microsoft research Asia as a researcher on data intelligence for cloud systems. She is now working as a research fellow to carry out research mainly in the medical NLP area.