CUHK Research: Changing the world

no real application.” He was disappointed but philosophical. “Typically, going from theory to application may take decades.” A killer application For applications of network coding to move beyond the laboratory, there were hurdles to overcome. For example, to mix data packets, it was necessary for intermediate nodes in the network to perform computation. But traditional networks used simple routers with little computing power. “Another hurdle was that we didn’t have an efficient algorithm for implementing network coding. In other words, the benefit offered was not sufficient to lure network operators to use a more elaborate way of transmitting information,” Professor Yeung recalls. “I began thinking we need a killer application – not one that merely improves network performance by 30-40% but one that is impossible without network coding.” He zeroed in on what is called the “multi- hop” curse in wireless communication. That is when packets of data are lost as they “hop” from one node to another because of noise or other channel interference. Accumulations of lost packets clog network transmission paths. Network coding can solve this problem by “opening” and mixing packets at intermediate nodes, encoding them, and then decoding them to reconstitute the original data using mathematical computation. To develop a solution for this killer application, Professor Yeung turned to former PhD student Yang Shenghao who was proficient in coding theory. “I gave him the multi-hop problem and within a relatively short time he was able to come up with a very elegant solution. That was Batched Sparse Codes (‘BATS’),” says Professor Yeung. This disruptive network coding technology reduces the rate of packet loss per “hop” while significantly increasing the rate of data transmission. Overturning folklore For a time, it was folklore in communication research that the best way to transmit information through a network was by generating bunches of data packets and routing them through nodes acting like postal relay stations. That was before CUHK information theorist Raymond Yeung questioned the assumption some 30 years ago. “My first question was whether data packets from different sources need to be mixed in the network,” he explains. He followed a hunch that by mixing data packets from two different sources, it was possible to transmit more. “It took me a week to come up with an example showing that this is actually the case.” That discovery intrigued Professor Yeung, who in 1991 left a premier US research lab to join CUHK’s then newly established Department of Information Engineering. At CUHK, he began his paradigm-shifting research. “At that point we didn’t have a very comprehensive theory for this phenomenon of mixing data packets inside the network. So, for the next seven or eight years I worked with individual collaborators here at CUHK and overseas. I knew it was going to lead to something very important.” Indeed, it led to network coding, a revolution in network communication, enabling more information to be transmitted through networks at a faster rate so that, for example, data can be downloaded faster from the Internet and video streamed with less delay. Breaking with convention, network coding introduced computation, or coding, at intermediate nodes within the network to increase throughput. “It’s a completely new way of t ransmi t t i ng i n fo rma t i on through a network,” says Professor Yeung. “Subsequently, we tried to find applications. However, after 10 years or so, there still was 21