Machine Learning has many definitions, but in my perspective, ability to acquire information from raw data is called Machine Learning. Applications of Machine Learning is well known in finance, image processing, natural language processing, driving etc. But what role does machine learning play in Communication Systems?
Our cities are becoming super smart and fully connected with autonomous devices. Everything around us is connected by means of Internet of Things (IoT). The fifth-generation communication system that is currently being deployed us aimed at providing faster, higher capacity transmission. There has been an exponential increase in amount of data that is being transferred. The amount of data that is being transferred per month is 80 Exabytes. That is approximately equivalent of transferring 10000000000 High-Definition movies. This comes as no surprise. The rapid development of AI, IoT, VR, Cloud technologies has led to a massive increase in data traffic. 5G infrastructure would reach its peak capacity by 2030’s bringing need for 6th generation communication system which can meet the demands for ever increasing high capacity, high speed communication system. And AI or ML is expected to play a major role.
At present, ML is communication networks is something that is unheard of or rare. This is because of the need for substantial improvement needed in infrastructure ML applications can be put into place. Also development of ML in communication networks requires multi-disciplinary skills that industry lacks.
ML loves data. Data is a fundamental building block that ML depends on. There are different types of networks, communication devices and equipment’s that produce different types of data. Proper infrastructure needs to be setup to collect and store all the available data which can then be further analyzed.
But what’s the use? How can ML be used in communication networks? As I mentioned earlier, data consumption is increasing and there is a need for low latency, long distance, high speed communication systems. Consider our todays networking system, the way packets are routed. The packets are routed based on traditional routing algorithms. All of which find the shortest and quickest route to reach the destination. But, what if the router has the capability to predict the network congestion that might occur? With large amounts of data that is produced, the shortest route may not be the quickest route always. The networks needs to be adaptive based on the current status and also needs to be future aware. The network should be able to predict the demands, network failures all of which needs further research. In the near future, more impactive ML applications in networking is expected to be developed. With ML, Software Defined Networks (SDN) becoming more mature, telecommunications is getting ready for ML driven applications.