Dynamic Features Adaptation in Networking: Toward Flexible training and Explainable inference

TL;DR

The paper tackles the need for AI that adapts to evolving feature exposure and non-stationary conditions in AI-native 6G networks. It introduces Adaptive Random Forests for dynamic feature adaptation and Drift-Aware Feature Importance (DAFI), which uses a Kolmogorov-Smirnov ($KS$-Test) drift detector to decide when to compute accurate but expensive SHAP explanations versus faster Mean Decrease in Impurity (MDI) estimates. The authors demonstrate that iterative ARF training yields increasing accuracy as new features are added and that DAFI substantially reduces FI runtime while improving consistency of feature attribution. The results on Electricity, Weather, and Network datasets indicate meaningful runtime savings (up to about 55% in some cases) and more stable explanations, supporting the viability of flexible, explainable AI for 6G networks.

Abstract

As AI becomes a native component of 6G network control, AI models must adapt to continuously changing conditions, including the introduction of new features and measurements driven by multi-vendor deployments, hardware upgrades, and evolving service requirements. To address this growing need for flexible learning in non-stationary environments, this vision paper highlights Adaptive Random Forests (ARFs) as a reliable solution for dynamic feature adaptation in communication network scenarios. We show that iterative training of ARFs can effectively lead to stable predictions, with accuracy improving over time as more features are added. In addition, we highlight the importance of explainability in AI-driven networks, proposing Drift-Aware Feature Importance (DAFI) as an efficient XAI feature importance (FI) method. DAFI uses a distributional drift detector to signal when to apply computationally intensive FI methods instead of lighter alternatives. Our tests on 3 different datasets indicate that our approach reduces runtime by up to 2 times, while producing more consistent feature importance values. Together, ARFs and DAFI provide a promising framework to build flexible AI methods adapted to 6G network use-cases.

Figures (4)

• Figure 1: DAFI: Drift-aware Feature Importance algorithm
• Figure 2: Performance of ARF for benign and attack network classifications with evolving Network KPIs - Network dataset Xavier2024 - epoch 0: start with features "mac_dl_mcs", "phy_ul_pusch_sinr", "phy_ul_pucch_sinr" - epoch 10: adding "phy_ul_pusch_rssi" - epoch 20: adding "mac_dl_cqi", "phy_ul_pucch_rssi" - $n_{\text{model}}=10$
• Figure 3: Influence of number of trees in the forest on Model and Feature importance algorithms Performance - Synthetic dataset with 80:20 train/test split - Metrics of algorithms are given for $n_\text{samples} = 100$ samples from the test dataset
• Figure 4: Illustration of the top-k metrics - Exact and Set matches