Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper tackles the challenge of cross-domain generalization in optical networks by introducing a novel joint contrastive and classification learning technique that captures stable, task-relevant relationships across heterogeneous domains. By simultaneously optimizing representation learning and task performance, the method effectively shapes the latent space to enhance model robustness. Experimental results on lightpath quality estimation show significant improvements over baseline methods, highlighting the approach's ability to adapt quickly with minimal fine-tuning.
Models can achieve robust performance across diverse optical network configurations with a novel joint learning approach that enhances adaptability and efficiency.
The robustness of machine learning techniques across heterogeneous network domains remains an open challenge in optical networks. Models trained on data from a specific topology or operational configuration often exhibit degraded performance when deployed in unseen networks. In this work, we address this challenge by proposing a representation learning technique aimed at capturing task-relevant relationships that remain stable across domains. The proposed technique is based on a novel joint contrastive and classification learning approach in which representation learning and task optimization are performed simultaneously, allowing both objectives to shape the latent space. Experimental results on a representative use case, namely, lightpath quality of transmission estimation, demonstrate the effectiveness of our approach compared to baseline approaches, and highlight its capacity for rapid adaptation, providing excellent performance even with limited fine-tuning.