Domain adaptation and transfer learning for failure detection and failure-cause identification in optical networks across different lightpaths [Invited]

Abstract

Optical Network Failure Management (ONFM) is a promising application of Machine Learning (ML) to optical networking. Typical ML-based ONFM approaches exploit historical monitored data, retrieved in a specific domain (e.g., a link or a network), to train supervised ML models and learn failures characteristics (a signature) that will be helpful upon future failures occurrence in that domain. Unfortunately, in operational networks, data availability often constitutes a practical limitation to the deployment of MLbased ONFM solutions, due to scarce availability of labeled data comprehensively modeling all possible failure types. One could purposely inject failures to collect training data, but this is time consuming and not desirable by operators. A possible solution is Transfer Learning (TL), i.e., training ML models on a Source Domain (SD), e.g., a lab testbed, and then deploy trained models on a Target Domain (TD), e.g., an operator network, possibly fine-tuning the learned models by re-training with few TD data. Moreover, in those cases when TL re-training is not successful (e.g., due to intrinsic difference of SD and TD), another solution is domain adaptation, which consists of combining unlabeled SD and TD data before model training. We investigate domain adaptation and TL for failure detection and failure-cause identification across different lightpaths leveraging real OSNR data. We find that, for the considered scenarios, up to 20 percentage points of accuracy increase can be obtained with domain adaptation for failure detection, while for failure-cause identification only combining domain adaptation with model re-training provides significant benefit, reaching 4-5 percentage points of accuracy increase in the considered cases.

Machine learning-based QoT estimation over optical wireless communication

“…Machine learning (ML) algorithms can be applied for fiber-optic communication and networks with both the physical layer and network layer [8], [9]. A regression algorithm is an analytical method that is used in optical communication.…”

A review of machine learning-based failure management in optical networks

“…For a lightpath, the typical failure types are bit error rate (BER) degradation, optical signal to noise ratio (OSNR) degradation, generalized SNR (GSNR) degradation, optical power drop, channel crosstalk, etc. Failure management on lightpath is a system-level management scheme concerning the overall system performance metrics [9,10].…”

Meta-learning-based failure localization with digital-twin-enabled multi-mirror models in optical networks

“…Several approaches have been proposed to overcome these issues. For example, transfer learning allows to transfer models trained in networks with sufficient data (i.e., the source domain) to networks with insufficient data (i.e., the target domain) [7]; or generative adversarial networks (GANs) can expand the number of available training samples [8]. However, transfer learning performance heavily depends on correlation between source/target domain, while GAN needs lots of samples to train generator.…”