Machine Learning for Spatiotemporal Sequence Forecasting: A Survey

PhD Qualifying Examination


Title: "Machine Learning for Spatiotemporal Sequence Forecasting: A 
Survey"

by

Mr. Xingjian SHI


Abstract:

Spatiotemporal systems are common in the real-world. Forecasting the 
multi-step future of these spatiotemporal systems based on the past 
observations, or, Spatiotemporal Sequence Forecasting (STSF), is an 
important and challenging problem. Compared to purely spatial data like 
images and purely sequential data like sentences and audios, 
spatiotemporal sequences not only contain information about 'what' and 
'when' but also provide information about 'where'. This makes them more 
comprehensive about the underlying system and also imposes new challenges 
to the machine learning community. Although lots of real-world problems 
can be viewed as STSF and many research works have proposed machine 
learning based methods for them, no existing work has summarized and 
compared these methods from a unified perspective. This survey aims to 
provide a systematic review of machine learning for STSF. In this survey, 
we first introduce the two major challenges of the problem: 1) how to 
learn a model for multi-step forecasting and 2) how to effectively model 
the spatial and temporal structures within the data. We then review the 
existing works for solving these two challenges, including the general 
learning strategies for multi-step forecasting, the classical machine 
learning methods for STSF and the deep learning methods for STSF. We also 
compare these methods and point out some potential research directions.


Date:			Tuesday, 17 January 2017

Time:                  	2:00pm - 4:00pm

Venue:                  Room 3494
                         Lifts 25/26

Committee Members:	Prof. Dit-Yan Yeung (Supervisor)
 			Prof. Nevin Zhang (Chairperson)
 			Dr. Qiong Luo
 			Dr. Raymond Wong


**** ALL are Welcome ****