Concept drift adaptation for multiple streams with temporal dependence
The School of EECS is hosting the following seminar:
Concept drift adaptation for multiple streams with temporal dependence
Speaker: Dr Yiliao Song, RMIT University
Host: Prof Helen Huang
Abstract: Real-time decision-making for streaming data is facing a common but challenging problem, concept drift. Namely, the data distribution changes over time. Classic machine learning schema might be invalid to obtain reliable prediction results for data streams that have the concept drift problem. This talk presents three solutions to address this issue. In addition, concept drift is discussed when it co-exists with temporal dependence in a data stream. In the end, this talk will drop a view on the future study of concept drift adaptation methodologies that facilitate reliable learning for multiple streams that are neither independent nor identically distributed.
Bio: Dr. Yiliao Song is a research fellow in Enterprise AI and analytics Hub at RMIT University. Her research interests focus on real-time decision support, especially learning methodologies of real-time prediction for non-stationary data streams. So far, she has 20 publications, with more than 900 citations. Her h-index is 13 and i10-index is 18.
About Data Science Seminar
This seminar series is hosted by EECS Data Science.