Search

Prediction of the quality of transmission (QoT) in optical communication networks is a critical task to ensure reliable transmission and optimized operation. Accurate QoT estimation is essential for enabling future reconfigurable optical networks, where lightpaths are set up based on performance predictions. There are three main categories of QoT estimation methods: exact analytical models offer high accuracy but require significant computational power, making them unsuitable for real-time applications, approximate formulas are faster to compute but sacrifice some accuracy, machine learning (ML) and deep learning (DL) methods potentially offer high accuracy with lower computational demands, making them promising candidates for real-time scenarios. DL has emerged as a powerful tool for modeling complex nonlinear relationships and has shown promise in QoT estimation. In this paper, we propose a DL-based classifier to predict whether a lightpath meets the required system performance, both in single-mode fiber (SMF) and few-mode fiber (FMF) links. Using synthetic datasets generated from the enhanced Gaussian noise (EGN) model, we compare DL-based classifier with ML-based classifiers, DL-based regressor, and closed-form models (CF-GN and CF-EGN). Results show that the proposed DL-based method outperforms the conventional ML-based method and closed-form models in both SMF and FMF links in terms of various classification metrics. Moreover the DL-based classifier outperforms the DL-based regressor in terms of complexity, while achieving nearly the same performance in both SMF and FMF links. This superiority stems from DL's capabilities: DL models can effectively capture the intricate interactions between various factors that influence QoT, DL networks can learn layered representations of the data, allowing for a deeper understanding of the underlying patterns, and DL models automatically extract relevant features from the data, eliminating the need for manual feature engineering. This work demonstrates the potential of DL for real-time and accurate QoT estimation in optical networks. The proposed DL-based classifier can be used to optimize link configurations, predict network performance, and enhance overall network reliability.

Biological invasions threaten water resources worldwide owing to interrelated biological and anthropogenic drivers of change. Though the role of people in the (un)intentional spread of aquatic invasive species is increasingly recognized as a priority, there are widespread assumptions about the importance of public awareness in explaining biosecurity behaviors. A disconnect exists between what people think and what they do—as articulated by the simple but powerful “knowledge-action” gap—that warrants attention to clarify how environmental social science research can most effectively advance invasive species prevention. Using survey data collected from aquarium hobbyists in the U.S. state of Illinois, we investigated the role of belief systems in shaping intentions to engage in biosecurity behaviors, such as purchasing only native species as pets. We also examined how beliefs were rooted in multiple forms of knowledge. Self-efficacy was the strongest driver of intended biosecurity behavior, whereas both perceived benefits and risk perceptions were instrumental in explaining hobbyists’ decisions. Beliefs were informed by two types of reported knowledge, which in turn, were influenced by a range of information sources. Our results suggest that a more nuanced conceptualization of public awareness is urgently needed given its critical but often misunderstood role in the formation of beliefs that influence participation in biosecurity behaviors. Knowledge of both biological invasions and biosecurity behaviors, in addition to information sources, needs to be distinguished in future research. These multiple forms of knowledge serve as the foundation for belief systems that activate behavior change among aquarium hobbyists. We also argue that science communication and outreach campaigns that highlight the risks of invasive species while strengthening self-efficacy and perceived benefits of prevention will be the most effective pathways for fostering hobbyist engagement in biosecurity behaviors.