Rahl – a new residual-based adaptive Huber loss function for CQI prediction in 5G networks

Abstract

The evolution of data consumption patterns, increasingly decentralized from human behavior, has exerted pressure on the transmission resources within the current mobile network infrastructure. In response to this challenge, Fifth Generation Networks (5G) have been standardized and globally deployed, incorporating high-frequency millimeter-wave links to meet the escalating demands for data transfer speeds.

5G signifies the fifth generation of wireless technology, distinguished by its enhanced data speeds, reduced latency, and expanded capacity in comparison to earlier iterations. This advancement holds the potential to transform communication, facilitating the deployment of cutting-edge applications like autonomous vehicles, augmented reality, and seamless Internet of Things (IoT) connectivity. Its extensive implementation is poised to revolutionize various industries and bolster connectivity within our increasingly digitalized environment.

However, the existing management approach for 5G networks is reactive, relying on user devices to calculate network parameters, which are then periodically transmitted to base stations. This reactive methodology introduces delays and network slowdowns, potentially jeopardizing the timely fulfillment of operational requirements.

In the realm of 5G networks, the Channel Quality Indicator (CQI) plays a pivotal role in adjusting modulation and coding schemes based on channel conditions, ensuring optimal data transfer rates and network performance. Recent research focuses on improving CQI estimation in 5G networks using machine learning. In this field, loss functions play a vital role, serving as a guide for training models and optimizing their performance. Two commonly used loss functions are Mean Squared Error (MSE) and Mean Absolute Error (MAE). Roughly speaking, MSE put more weight on outliers, MAE on the majority.

Here, we argue that the Huber loss function is more suitable for CQI prediction, since it combines the benefits of both MSE and MAE. To achieve this, the Huber loss transitions smoothly between MSE and MAE, controlled by a user-defined hyperparameter called delta. However, finding the right balance between sensitivity to small errors (MAE) and robustness to outliers (MSE) by manually choosing the optimal delta is challenging.

To address this issue, we propose a novel loss function, named Residual-based Adaptive Huber Loss (RAHL). In RAHL, a learnable residual is added to the delta, enabling the model to adapt based on the distribution of errors in the data. Our approach effectively balances model robustness against outliers while preserving inlier data precision.

The widely recognized Long Short-Term Memory (LSTM) model is employed in conjunction with RAHL, showcasing significantly improved results compared to the aforementioned loss functions. The RAHL has enhanced prediction accuracy for datasets A, B, and C by approximately 11%, 14%, and 0.3%, respectively, compared to the Huber loss function; by 22%, 23%, and 17%, respectively, compared to the MSE loss function; and by 5%, 10%, and 5%, respectively, compared to the MAE loss function. The obtained results affirm the superiority of RAHL, offering a promising avenue for enhanced CQI prediction in 5G networks.