INESC TEC researchers win international competition of Artificial Intelligence for 5G networks

A team of researchers from INESC TEC, in collaboration with the Polytechnic Institute of Leiria, won the second edition of the international competition AI/ML in 5G Grand Challenge, organised by the International Telecommunication Union (ITU) – within the scope of the development of more efficient artificial intelligence algorithms, namely deep learning algorithms for 5G communications networks.

This acknowledgment was based on the work developed in a neural network for classifying radio signal modulations, which is computationally efficient and allows the classification of modulations in real time, particularly in devices with limited resources. “This solution distinguished itself by reaching the minimum level of precision required by the competition organisers, having been the most computationally efficient model, among the solutions submitted to the competition”, said Daniel Granhão, researcher at INESC TEC’s Centre for Telecommunications and Multimedia (CTM).

This solution is available in open source and aims to improve the efficiency of radio-based wireless connections, as well as the migration of the implementation of Artificial Intelligence (AI) solutions from the cloud to the edge, i.e., closer to the data source. “This solution is computationally efficient, and it can be integrated into devices on the edge, with more limited resources than cloud devices, thus minimising some of the problems inherent to the implementation in the cloud itself”, added Daniel Granhão.

Thanks to this acknowledgement, the team of researchers (called BacalhauNET) composed of INESC TEC researchers Daniel Granhão, Guilherme Carvalho and Tiago Gonçalves, as well as José Nuno Grácio Rosa, from the Polytechnic Institute of Leiria, obtained recognition from the international community, the opportunity to publish their work in an ITU scientific journal, and a prize of 5000 CHF (approximately 4800 thousand euros).

This edition of the ITU AI/ML 5G Grand Challenge featured 16 challenges and brought together over 1600 participants from 82 different countries. This competition consists of two parts: in the first, the jury selects the two/three best teams for each challenge; in the second, the best team among the 16 open challenges is selected as the winner of the Grand Challenge. The BacalhauNET team won the Grand Challenge with a solution to the “Lightning-Fast Modulation Classification with Hardware-Efficient Neural Networks” challenge.

The INESC TEC researcher mentioned in this news piece is associated with FCT.