Towards the Uncertainty Quantification of AI-Models in 6G Networks

The evolution towards the 6<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sup> generation of telecommunication systems blazes the trail for the advent of a plethora of innovative technologies that aim to radically change the current telecommunication systems. The new generation of networks is challenged to serve highly demanding applications and services that are hosted in a substantial number of devices. The complexity of the new telecommunication ecosystems has designated the adoption of AI solutions as inevitable in order to conduct fundamental network operations such as resource management and orchestration. However, the AI models are not always able to provide certain estimates, and they may suffer from under- or over- confidence in their predictions. With a view to tackling this issue, in this work, we employ an innovative framework aiming to quantify the uncertainty in the estimates predicted via two DL models, i.e., the LSTM and the 1D-CNN, which are trained with telecommunication data. It is shown that the uncertainty quantification procedure is a strong aspect in designing trustworthy AI models since it increases the probability that the actual predictions are within the interval of the model predictions.