Bridging the Gap Between Simulated and Real Network Data Using Transfer Learning

2025-10-01

AI Summary

Network machine learning models trained on simulated data suffer from degraded prediction accuracy in real-world deployments due to domain shift between simulation and reality. Method: This paper proposes a lightweight hybrid transfer learning framework that bridges the simulation-to-reality gap using only a small number of real-network samples. Built upon RouteNet-Fermi, it jointly leverages OMNeT++-generated synthetic traffic data and real traffic traces collected from a custom testbed for efficient fine-tuning. Contribution/Results: The approach significantly reduces reliance on large-scale, labeled real-world data—addressing critical data scarcity in rare but pivotal scenarios (e.g., network failures). Experimental evaluation on packet delay prediction shows up to an 88% reduction in mean absolute percentage error (MAPE). With merely 10 real-world scenario samples, MAPE decreases by 37%; with 50 samples, the reduction reaches 48%. These results demonstrate substantial improvements in model robustness and generalization capability for practical deployment.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Planning, Routing, and Scheduling: Planning/Scheduling and Learning
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Web data generation and simulation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Machine Learning (ML)-based network models provide fast and accurate predictions for complex network behaviors but require substantial training data. Collecting such data from real networks is often costly and limited, especially for critical scenarios like failures. As a result, researchers commonly rely on simulated data, which reduces accuracy when models are deployed in real environments. We propose a hybrid approach leveraging transfer learning to combine simulated and real-world data. Using RouteNet-Fermi, we show that fine-tuning a pre-trained model with a small real dataset significantly improves performance. Our experiments with OMNeT++ and a custom testbed reduce the Mean Absolute Percentage Error (MAPE) in packet delay prediction by up to 88%. With just 10 real scenarios, MAPE drops by 37%, and with 50 scenarios, by 48%.

Problem

• Reducing accuracy gap between simulated and real network data
• Minimizing real data requirements for machine learning models
• Improving packet delay prediction in network deployments

Innovation

• Uses transfer learning to combine simulated and real data
• Fine-tunes pre-trained models with small real datasets
• Reduces packet delay prediction error by up to 88%