Page 80 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Engineering
Department of Computer Science & Engineering

Data visualization has been prevalent around the world for years. It is extremely useful in the field of analyzing specific
objects’ changes over a period of time. In data visualization, we typically adapt tables and graphics to make data
accessible and easy to be compared. And we also gather statistics in an interval and process it to see how its
corresponding object changes during this interval. HKUST is a technological university accommodating over 10000
students and staff who busily cross through various zones. Therefore, as for our project, we choose people in HKUST to
see how they move around in the university every day. How people’s locations vary with time and which approach they
select interest us a lot. After studying it, we develop a visual web page to demonstrate these contents and draw some
conclusions about people’s daily lifestyle in HKUST. In a long run, this system can undoubtedly be applied to more
general cases and the result will be more influential.

Efficient Algorithms to Process Gigapixel Images
Supervisor: SANDER Pedro / CSE
Student: SUGIMOTO Ryusuke / COSC Course: UROP1000, Summer

The final target of this project is to generate video loops of gigapixel panorama from several still pictures. In order to
realize this, we will first identify what kind of objects there are in the image and to render the objects using the existing
rendering techniques. Therefore, we need to implement several different rendering techniques. We will start from
water surface to inspect the validity of our idea since water surface is a common non-static object in our daily scene. As
a progress report, we will discuss the basic rendering techniques available today and the techniques to render water
surface.

Heterogeneous Information Network Mining

Supervisor: SONG Yangqiu / CSE
Student: CHEN Sixuan / COMP Course: UROP1100, Spring
UROP2100, Summer

In this report, I would firstly present the overall ideas and objectives with a detailed plan for this project of improving
the current meta-path based FMG model. Precisely, Attention based method is being examined in combine with
deepFM model, and metapath2vec would be implemented on it in place of the current embeddings. The overall
introduction is followed by a specific explanation of Attention method, especially focusing on meta-path based co-
Attention model (MCRec) and aspect-level deep CF model (NeuACF). Next, some parts of hands-on implementations in
code level and my learnings will be covered. Furthermore, a brief introduction on metapath2vec model would be shown,
and the report is concluded with future works on code implementations, and final realization of the entire framework
that is fully connected through deep neural network to automatically learn all the parameters end-to-end, that finishes
the whole project.

Supervisor: SONG Yangqiu / CSE
Student: WANG Yili / COMP Course: UROP1100, Spring
UROP2100, Summer

Stack Overflow is a platform for users to ask and answer question on all the fields of computer science. According to [1],
Stack Overflow has more than 900,000,000 registered users up to August 2018 and more than 16,000,000 questions up
to mid2018. There is a large amount of data embedded in Stack Overflow that we can use to optimize user experience.
We are currently working on a project to transfer data sets of Stack Overflow to a heterogeneous information network

70