Pioneering Work and Study in Data Science

Dec 24, 2021


Many are no stranger to the concept of big data in today’s world. But few may be aware of the diverse job opportunities and burgeoning demand for data scientists and engineers. Beginning from the current academic year, the Faculty of Engineering of the University of Hong Kong (HKU) has offered data science as an option of a minor field of study to all engineering undergraduates. Students can choose it as major from 2022 academic year onwards and a new Bachelor degree in Data Science would be launched.

Data science-related skills have become more and more important, says Professor George Q. Huang, chair professor and head of Department of Industrial and Manufacturing Systems Engineering says, “Students will need to be well-versed in mathematics, statistics and computer science to be fully trained data scientists. Data science degrees are increasingly being offered by top universities in the world.”

As a rising number of students have based their dissertations on data-related fields, and with a critical mass of scientists in the areas, the Faculty is seeking to not just accelerate related research but also to nurture graduates who will become future data-driven leaders.

Three special clusters – in data science, environment and health – have been set up by the faculty to achieve synergy through research collaborations.


Professor George G.Q. Huang

Professor Huang is director of the Data research cluster, which is part of the newly-established Institute of Data Science (HKU-IDS), intended to provide an open platform for multi-disciplinary, cross-departmental research, covering pivotal areas such as machine learning, data science, and artificial intelligence. “The three clusters have become more and more data-driven,” says Professor Huang.

Expansion on multiple fronts

He expects a rapid expansion in the scope, scale and depth of data technology research and application in the coming years. That will help meet the rising need from businesses and organisations for support in areas other than data collection alone.  Examples include autonomous smart systems; advanced systems for data cleansing; data storage and retrieval; data presentation.

“There will be more insightful, and impactful outcomes as data technology continues to develop,” says Professor Huang. “In the past few years, professors have found sets of data, like huge treasures for their research; that’s exciting but they spend quite a proportion of their time cleansing the data. Some of the retrieved data may be usable for research. They end up spending much time on the tedious task of data cleansing. In the future businesses will ask data scientists to specify the data to be collected, the ideal scope etc.”

He said the next stage of data science development will take a life cycle approach, making sure that the right data is collected over a right period of time in the right format for analysis, and that the insight drawn from analysis can be stored. “Students and companies lack such knowledge now; with the right tools and software used for each step in the data collection process, successful data analysis can be achieved.


Social Value Generated from Data Management

Professor Reynold C.K. Cheng

Professor Reynold C.K. Cheng of the Department of Computer Science has for years used data analytics to improve people’s lives.

His project named "HINCare: A Heterogeneous Information Network for Elderly-Care Helper Recommendation" won the 2021 Faculty Knowledge Exchange Award. HINCare is an AI platform that enables timebanking, in which a volunteer earns time credits for performing a service (e.g. taking an elder to a clinic) in return for a service provided by another volunteer.

Downloadable from Apple and Google Play Store, HINCare’s Heterogenous Information Network stores the relationship information among more than 1,000 elders, helpers, and has been used by leading NGOs.

The network involves graph data and highlights the relationship between different objects and entities. Though recording relationships in an unstructured manner, a wider use of graph data system can be expected for the future, said Professor Cheng. “They can capture a lot of different things, all connected with each other like a complex network,” he explained.

In HINCare, by resorting to AI and big data, the network was used to find out the best match between an elderly person and a potential helper, without relying on what is known by an NGO or a social worker.  "Our project integrates Big Data, AI, and gerontology solutions," said Professor Cheng.  “The gigantic network from this project will be useful to gerontology researchers and policymakers.”

The same kind of network is also being used by a mainland television manufacturing company, to help them match viewers with videos believed to appeal to them. Again this is made possible by a graph database containing viewers’ past viewing habits.

Creating such database is a complex task, and requires both knowledge and stamina in carrying out repeated trials. “There are many software tools available but you need to know the right tools to use, before you can discover insight from the data. It’s not easy at all,” said Professor Cheng. “You need to have basic concepts of data mining, neural networks, machine learning methods, statistical learning etc.”

MTR passengers’ Travelling Patterns

An expert in large-scale data management, he has for the past two years been engaged in a major project related to the pandemic. Following an agreement signed with MTR Corporation, his team is keeping track of the relationship between Covid1-9 and MTR passengers’ travelling modes, eg. how an outbreak in a particular district affects passengers’ travelling modes. Anonymous, gigabytes of data were given to Professor Cheng by the company regularly.

He said one challenge is deciding on how to use the data, what sorts of questions to be asked. This could trigger some interdisciplinary research. Indeed, his collaboration with the Department of Architecture unearthed data on important concepts such as ‘familiar strangers’.

His team is currently building up a data analysis platform so analysis can be undertaken by various researchers for various purposes in the future. HKU’s agreement with MTRC involves developing possible solutions to enhance railway operation and maintenance engineering forecasting, image and data visualization, and big data analytics.

A good platform with lots of built-in visualisation facilities makes it easier for researchers to understand hidden patterns or trends. Already Professor Cheng has assembled a team of students to develop the visualization facilities.

“We will demonstrate the platform at the Tam Wing Fan Innovation Wing Two at the end of next year. Hopefully more researchers can use it for analysis later.”


Smooth Flow of Data in Powerful Transmission

Dr Huang Kaibin

Mobile internet is an indispensable part of modern day lives and ensuring the smooth transfer of data is undoubtedly both timely and critical.

Research in the area has grown as various spheres of life are now reliant on data stored in mobile phones, sensors and other devices. That is also the focus of the Associate Head of the Department of Electrical and Electronic Engineering, Singapore-educated Huang Kaibin. “The Internet of things is the new paradigm of computing,” he said. “People rely on tens of billions of devices to do computing; we are in a data economy where chips have become very cheap.”

Dr Huang is working on making sure that data flow is streamlined and less congested while numerous networks are operating in our surrounding environment, transmitting data from all directions, vertically or horizontally.

Increasingly, data are used for wide-ranging purposes from security surveillance to road traffic control. “Various procedures are controlled by wireless networks; for example, a robot operating on a patient in another city while being controlled or watched by a doctor in a distant location. The robot follows the doctor’s actions or instructions but we need to make sure there is very low latency, otherwise it can cause death. My research is on how to link up these computers.”

Powering up Networks

In particular, he specializes in a power beacon network that can recharge interconnected devices on a large scale using a dedicated station. Devices are given wireless access to power through the use of microwave. The network also serves the purpose of training up artificial intelligence (AI) by gathering complete data, just as gas, oil or other energy is collected in a pipe line.

Nowadays, he says, data are like fuel or key power source to companies. “Recently we built a network allowing machines to learn from data from different sources, on users’ behaviours, happenings in the surrounding environment, etc., knowledge data collected from tens of thousands of devices. We use the data to train machines to be more intelligent and learn faster,” explained Dr Huang.

To cope with the exponential growth in data traffic, as conveyed messages today consist of videos, virtual/augmented reality images and soon holograms, i.e. 3D images as well, he is also looking at ways to transmit data in the most efficient manner and to extract only the key messages for transmission. “The traditional way of compressing all the information into bits is placing a heavy burden on networks.”

Inspired by theories

While an undergraduate, he was initially studying civil engineering but his passion changed after he transferred to the National University of Singapore about 20 years ago. He switched to electrical and electronic engineering, intending to study circuits and optical networking. After attending an eye-opening course on wireless communication, he gave up cable communication in favour of wireless communication. It had to do with two teachers at the university’s Centre for Wireless Communications, who wrote theories on the blackboard which captivated the young Huang. He found them ‘elegant and inspiring’.

“I just wanted to be like them,” he recalled. “They were passionate about 3G technology and used fancy terms that caused me to change my area of interest.” He later finished his Ph.D. studies in the area at the University of Texas at Austin.

His task now is the development of applications for the 5G, 6G world. He foresees precise wireless communication tools being applied in more and more aspects of human lives, including remote surgery, healthcare, AI, manufacturing and autonomous driving. “Mine is an integrated research looking more at the integration of AI and wireless communication,” he said.


Robots as a part of everyday life

Dr Pan Jei

Robots are commonly seen today but they could be used much more widely and for a greater variety of purposes in the days ahead.

That is the vision of Pan Jia, Assistant Professor in the Department of Computer Science, who for more than a decade has conducted research on robotics and artificial intelligence. He believes autonomous robots are possible through the deployment of intelligent algorithms, sensors and machines.

“Now robots now are mostly manually controlled by humans, but we hope they can learn from data so they can perform complicated jobs, for example in factories, by learning from data collected from workers or practitioners,” said Dr Pan, who began his robotics research while a PhD student at the University of North Carolina, prior to being a post-doctoral fellow at the University of Berkeley.

Enabling robots to learn can be very useful for the industrial sector, he said, by allowing factories to make customized products, catering to the demand of even just a small clientele. “Robots can take up those orders and so the assembly works according to specific requirements.”

Human-robotic interactions

One of his research outputs is the creation of a soft tactile sensor enabling robots to do more complicated tasks, and have more interaction with humans. Assisted by the input of data and deep learning, the sensor mounted at the fingertip of a robot can help it accomplish challenging tasks such as stably grasping fragile objects, dexterous manipulation, or checking the quality of garments produced. “Currently most sensors are like cameras, mainly used to scan objects,” Dr Pan explained. “But in actual work, there may be a need to feel friction, or in factories, workers may want to check the quality of work, whether there are holes in garments made. Instead of touching the garment surface with their hands or using cameras, they can use sensors. “

Dr Pan agrees that in certain tasks, like pouring tea for an elderly person, numerous trials would be needed to ensure safety and accuracy. But he envisages plentiful functions robots could be capable of, such as performing surgeries through data collected from sensors attached to the hand of a doctor monitoring the surgery remotely.

Yet right now, there are still risks involved with the procedure, which fuels the need for further research. “Any chance of delay in medical procedure could have serious consequences. That is why we need to enable robots to have better intelligence or humans to provide high-level guidance.”

His prime goal is to create human-like robots through machine learning and artificial intelligence. Achieving that represents a major breakthrough since the application of AI has thus far been mostly limited to the virtual world. Advancement in research and development could allow robots to serve humans on a much wider scale. 

“There will be much more drastic impact on human lives if AI can go into the physical world,” said Dr Pan. “We are optimistic that robots can do many useful tasks in the coming 20 years. Every 5 years there will be new applications. For example, we may be able to control robots at home remotely.”