Abstract
Connecting people offers opportunities to build communities of any size and consequently, brings the world closer together. Conventionally, the connectivity has happened through traditional radio-frequency communication methods. However, the ever-increasing demand for higher data-rate communications and the explosion of advanced wireless applications such as virtual reality, augmented reality, and internet of things, reduce the effectiveness of these method. Therefore, developing next-generation technologies, such as learning-based communication systems, that can satisfy the large data and ultra-high rate communication requirements would be of interest.
To address the challenging problem of connectivity, our research focuses on developing a learning-based framework for the next-generation communication systems. These systems can proactively adapt their communication and networking strategies to the dynamic of the environment, thereby maximizing their end-to-end performance in terms of data-rates, energy-efficiency, and link-reliability. Toward this goal, first information-theoretical tools are used to establish the fundamental limits (including bounds on the end-to-end performance). These performance limits are the keys for building reliable and efficient systems. Then, powerful machine learning techniques, such as deep learning, are employed for the implementation of such systems. In particular, a simple and cost-effective system with near-optimal performance can be implemented by merely taking off-the-shelf deep learning models, applying them to communication design problems, and tuning them based on the training data.