Home > Academic Announcements > (Nov. 24) Heterogeneous Features Integration in Deep Knowledge Tracing

(Nov. 24) Heterogeneous Features Integration in Deep Knowledge Tracing

Last updated :2017-11-23

Topic: Heterogeneous Features Integration in Deep Knowledge Tracing
Speaker: Dr. Haiqin Yang
(Hang Seng Management College)
Host: Dr. CHEN Chuan
Time: 16:00-17:00, Friday, November 24, 2017
Venue: 101 Lecture Hall, School of Electronics and Information Technology, Guangzhou East Campus, SYSU

Abstract:
Knowledge tracing is a significant task in educational data analytics with the goal to automatically trace students’ knowledge states by analyzing their exercise performance. Recently, Deep Knowledge Tracing (DKT) model has been proposed and shown significant improvement to solve this task. However, the original DKT model assumes that students will gain proficiency when practicing more, which overlooks the scenarios that some students may conduct exercises mindlessly. To overcome this issue, we propose to include the rich heterogeneous features into the DKT model implicitly. We demonstrate that by representing the predicted response and the true response into only a four-bit binary code and combining it with the original one-hot encoding as the input of a Long Short-Term Memory (LSTM) model, we can gain more information and improve the predicting accuracy on students’ exercise. This paper was selected in the best student paper awards finalists of ICONIP 2017.

About the speaker:
Dr. Haiqin Yang is currently an Assistant Professor at Department of Computing and the Associate Director of Deep Learning Research and Application Centre at Hang Seng Management College. He received the B.Sc. degree in computer science from Nanjing University, China, and the M.Phil. and Ph.D. degrees from The Chinese University of Hong Kong. His research interests include machine learning and big data analytics. In these areas, he has published two books, over fifty refereed journal and conference papers, which are scored about 2000 citations with an H-index of 20 from Google Scholar and an H-index of 15 from Scopus. He has initiated and successfully organized five international workshops in the topics of Scalable Data Analytics and Scalable Machine Learning. He has served the editorial board of Neurocomputing, a senior program committee member of AAAI (2017, 2018), IJCAI (2017, 2018) and a program committee member of international conferences such as AISTAT, ACML, CIKM, IEEE BigData, and a reviewer for over ten prestigious journals.