Cross-Environment Transfer Learning for Robust mmWave Path Loss Modeling in 6G Wireless Networks

Accurate millimeter-wave (mmWave) path loss modeling is critical for 6G and beyond wireless networks but remains challenging due to environment-dependent propagation effects and limited measurement data. This paper proposes a novel cross-environment transfer learning framework that combines parameter-based inductive transfer learning with structured pruning to enable robust and dataefficient path loss prediction across heterogeneous scenarios. The framework first trains a deep neural network on a data-rich source domain and selectively fine-tunes higher layers on a sparse target domain, preserving transferable propagation features while reducing computational cost. Structured pruning further removes low-importance neurons, resulting in a 38% reduction in trainable parameters with negligible accuracy loss. Experimental evaluation on multiple public mmWave datasets demonstrates that the proposed approach reduces root mean squared error (RMSE) by up to 15% and mean absolute error (MAE) by up to 12% compared to baseline neural networks, outperforming existing transfer learning strategies in both accuracy and efficiency. These results indicate that the hybrid transfer learning and pruning strategy provides an effective, scalable, and computationally feasible solution for cross-environment mmWave path loss modeling, with potential extensions to sub-terahertz networks and adaptive online learning scenarios.