Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Self-Supervised Transformer Architecture for Change Detection in Radio Access Networks

Igor Kozlov, Dmitriy Rivkin, Wei-Di Chang, Di Wu, Xue Liu, Gregory Dudek

TL;DR

This work tackles the challenge of detecting lasting changes in Performance Measurement data from Radio Access Networks. It introduces TREX-DINO, a self-supervised transformer-based framework derived from DINO, adapted to multivariate time series with 1D convolutions and multi-crop augmentations, to produce stable, discriminative representations for change detection. Evaluated on a proprietary, SLS-simulated dataset with thousands of cell-level PM metrics, TREX-DINO outperforms non-ML and ML baselines (Binseg and TIRE) in F1 score and PR AUC, demonstrating strong scalability and generalizability with minimal expert tuning. The approach enables scalable monitoring of RAN operations, offering metric-level insights that can be aggregated into higher-level events for operators.

Abstract

Radio Access Networks (RANs) for telecommunications represent large agglomerations of interconnected hardware consisting of hundreds of thousands of transmitting devices (cells). Such networks undergo frequent and often heterogeneous changes caused by network operators, who are seeking to tune their system parameters for optimal performance. The effects of such changes are challenging to predict and will become even more so with the adoption of 5G/6G networks. Therefore, RAN monitoring is vital for network operators. We propose a self-supervised learning framework that leverages self-attention and self-distillation for this task. It works by detecting changes in Performance Measurement data, a collection of time-varying metrics which reflect a set of diverse measurements of the network performance at the cell level. Experimental results show that our approach outperforms the state of the art by 4% on a real-world based dataset consisting of about hundred thousands timeseries. It also has the merits of being scalable and generalizable. This allows it to provide deep insight into the specifics of mode of operation changes while relying minimally on expert knowledge.

Self-Supervised Transformer Architecture for Change Detection in Radio Access Networks

TL;DR

This work tackles the challenge of detecting lasting changes in Performance Measurement data from Radio Access Networks. It introduces TREX-DINO, a self-supervised transformer-based framework derived from DINO, adapted to multivariate time series with 1D convolutions and multi-crop augmentations, to produce stable, discriminative representations for change detection. Evaluated on a proprietary, SLS-simulated dataset with thousands of cell-level PM metrics, TREX-DINO outperforms non-ML and ML baselines (Binseg and TIRE) in F1 score and PR AUC, demonstrating strong scalability and generalizability with minimal expert tuning. The approach enables scalable monitoring of RAN operations, offering metric-level insights that can be aggregated into higher-level events for operators.

Abstract

Radio Access Networks (RANs) for telecommunications represent large agglomerations of interconnected hardware consisting of hundreds of thousands of transmitting devices (cells). Such networks undergo frequent and often heterogeneous changes caused by network operators, who are seeking to tune their system parameters for optimal performance. The effects of such changes are challenging to predict and will become even more so with the adoption of 5G/6G networks. Therefore, RAN monitoring is vital for network operators. We propose a self-supervised learning framework that leverages self-attention and self-distillation for this task. It works by detecting changes in Performance Measurement data, a collection of time-varying metrics which reflect a set of diverse measurements of the network performance at the cell level. Experimental results show that our approach outperforms the state of the art by 4% on a real-world based dataset consisting of about hundred thousands timeseries. It also has the merits of being scalable and generalizable. This allows it to provide deep insight into the specifics of mode of operation changes while relying minimally on expert knowledge.
Paper Structure (11 sections, 7 equations, 3 figures, 1 table)

This paper contains 11 sections, 7 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Anomaly Detection in Cellular Networks
  4. Change Detection
  5. Change detectors
  6. Binseg
  7. TIRE
  8. TREX-DINO
  9. Dataset
  10. Performance Evaluation
  11. Conclusion

Figures (3)

  • Figure 1: A schematic illustration of an SLS environment that consists of 7 sites, which are serving a number of active and idle User Equipments (UEs).
  • Figure 2: Examples of discrete (top) and continuous (bottom) timeseries data created using SLS. The color coding indicates distinct system parameters sets.
  • Figure 3: Differential comparison of CPDs' performance as a function of normalized detection threshold.