A Survey on Transfer Learning for Urban Spatio-temporal Machine Learning

PhD Qualifying Examination


Title: "A Survey on Transfer Learning for Urban Spatio-temporal Machine 
Learning"

by

Mr. Yilun JIN


Abstract:

Urban computing has become an active research field in machine learning. 
Due to the developments of data infrastructures, such as positioning 
systems, wireless sensors, etc., large-scale spatio-temporal urban data 
are increasingly available, which motivates machine learning on urban 
spatio-temporal data, i.e. spatio-temporal machine learning (STML). STML 
has achieved success in various applications that facilitate citizens’ 
lives, such as transportation, environment, etc. However, it is common for 
cities to have insufficient data, as new cities are continually being 
built, and some cities may not have advanced data infrastructures. Under 
these conditions, STML methods suffer from performance degradation. 
Transfer learning, which borrows knowledge learned from a data-rich domain 
to a data-sparse domain, would help alleviate the lack of data in cities.

In this article, we present a survey on efforts to use transfer learning 
in the context of urban STML. We begin the survey by presenting key 
concepts in both urban STML and transfer learning. We then categorize 
existing literature on urban spatio-temporal transfer learning (STTL) into 
three classes: spatial, temporal, and cross-modality transfer learning, 
and introduce related works accordingly. We conclude the survey by 
pointing out unresolved challenges in urban STTL.


Date:			Wednesday, 18 August 2021

Time:                  	2:00pm - 4:00pm

Zoom meeting:
https://hkust.zoom.us/j/98641439916?pwd=cElkSDB3T0dUNDcycHdoSjVzdy9lQT09

Committee Members:	Prof. Qiang Yang (Supervisor)
 			Dr. Kai Chen (Supervisor)
 			Dr. Wilfred Ng (Chairperson)
 			Prof. Dit-Yan Yeung


**** ALL are Welcome ****