WIDCOMP - Wireless Digital Computations

There are major expectations in academia, industry, and society that wireless Internet of Things (IoT) and Machine Learning (ML) will revolutionize any digital technology, creating a new trillion-euro economy.

Envisioned services include autonomous vehicles, humanoid robots, virtual medical doctors, and pervasive augmented and virtual reality.

However, when we use ML over wireless, we face major challenges in terms of computation, bandwidth, scalability, privacy, and security.

The reason is that, on the one side, current wireless communication networks are not intended to support ML services, and on the other side ML theory is not conceived to deliver algorithms that can efficiently run over wireless links. Specialized wireless protocols for ML computations are emerging to overcome these problems.

Recently, it has been noted that the superposition of radio waves can be exploited for computational purposes to perform complex computations over-the-air.

However, the current-state of over-the-air computations is limited to analog amplitude modulations, with digital over-the-air computations facing a complete lack of understanding of fundamental methods and poor algorithm performance.

This research project aims to address these gaps and contribute to establishing a new theory for digital over-the-air computations and its application to ML, potentially producing a leap forward in the interdisciplinary domain of wireless communications for ML computations.