时间：1月11日 13:30
地点：理工楼212会议室
In recent years, AI has demonstrated super-human performance in image processing, speech analysis, natural language processing and many more. Unfortunately, existing state-of-the-art models lack transparency and interpretability, which impedes AI and Machine Learning from be applied in many traditional fields such as the medical, finance and politics. Consequently, the interpretability of AI has been widely concerned by academics and industry and is expected to become an emerging and promising direction. Although some studies have largely advanced the Interpretability rsearch landscape, they still lack causal interpretations that is needed for true understanding by humans.
This talk will start with the review of state-of-the-art in interpretable AI, which devotes to exploit the black-box nature of the AI models for justifying the model reliability. Then we report our recent research on causal recommendation and causal learning for interpretable AI via causal graph, prior privilege information and transfer learning. The talk ends up with discussions on some open questions and promising and directions towards high-quality Interpretable AI.
Bio: Dr Guandong Xu is a Professor at School of Computer Science, University of Technology Sydney, specialising in Data Science, Recommender Systems, Web Data Mining, and Predictive Analytics. He has published three monographs and 250+ papers in leading journals and conferences. He leads Smart Future Research Centre and Data Science and Machine Intelligence Lab at UTS. He is the assistant Editor-in-Chief of World Wide Web Journal and serving in editorial board or guest editors for several international journals. He has received a number of Awards from academia and industry, e.g. Top-10 Australian Analytics Leader Award and Australian Computer Society Disruptors Award.
