+1 Recommend
1 collections
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      Bridging the fields of optics and communications, statistics, signal processing and machine learning


      Science Impact, Ltd.

      Read this article at

          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.


          Professor Alan Pak Tao Lau, from the Hong Kong Polytechnic University, is combining optics with statistics and machine learning to boost the signals of optical communication networks beyond the current limits of the technology. According to Lau, ‘The emerging dominance of mobile devices, data centres, cloud computing, 5G wireless and IoT applications impose unprecedented speed requirements that are not met by this generation of technology.’ The solutions to some of the issues currently facing optical communications are often ones of fundamental physics. Again, according to Lau, ‘in contrast to data centres, 5G and IoT applications in which cost is a prime concern, long-haul transmissions are limited by the underlying physics of how electromagnetic fields of light interact with the fibre molecules.’ Therefore, the solutions require new theoretical advances to push the limits of the signals’ ability to travel. Specifically, for long-haul transmissions, city-to-city or cross-country distances, the light travelling through the optical fibres undergoes the Kerr effect in which the refractive index of the fibre actually changes with signal power, which in turn leads to signal distortions that is difficult to compensate for. This is just one problem that Lau and colleagues are looking to tackle. The first step in these investigations is to devise a theoretical solution based on physics and engineering and then to test it. To combat the Kerr effect, for example, the team has been experimenting with a technique called ‘nonlinear frequency division multiplexing’. ‘The mathematics of nonlinear frequency division multiplexing showed that signals transmissions using such a nonlinear frequency are immune to the fibre Kerr effects and hence, this is a major fundamental breakthrough that holds tremendous promise,’ says Lau. This solution is an excellent example of how interdisciplinary the field has become, requiring the talents of traditional physicists, engineers and now specialists in mathematics, analytics and even machine learning as well. ‘Machine learning represents a next step in signal processing beyond conventional statistical signal processing,’ says Lau. ‘Machine learning is fundamentally a data science,’ explains Lau, who adds that, ‘optical communications and networks are generating trillions and trillions of data packets per second.’ The advancement of optics and digital signal processing increased the amounts of data possible and made significant impacts on point-to-point optical communications which, according to Lau, ‘in turn lay the foundations for more advanced algorithms such as machine learning to come into the field.’ Machine learning is an extremely powerful way to find algorithms that can process and handle large amounts of data and certain problems in optical communications and networking are still incredibly difficult to mathematically model, making machine learning the ideal technique for finding these solutions. Recently, we partnered with the University of Bristol to demonstrate how machine learning can learn the network conditions and adapt its traffic in real-time over a deployed link from Bristol to Froxfield. This is an important step towards intelligent software-defined networks that is envisioned to be the key technology for optical networks in the next decade.

          Related collections

          Author and article information

          Science Impact, Ltd.
          December 12 2018
          December 12 2018
          : 2018
          : 9
          : 97-99
          © 2018

          This work is licensed under a Creative Commons Attribution 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

          Earth & Environmental sciences, Medicine, Computer science, Agriculture, Engineering


          Comment on this article