Organizers
Tingjun Chen, Duke University, USA
Dan Kilper, Trinity College Dublin, Ireland
Camille Delezoide, Nokia Bell Labs, France
Background, Motivation, and Objective
In recent years, machine learning (ML) has proven to be a transformative tool in modeling and understanding complex real-world data and systems. Its applications have revolutionized fields such as natural language processing, image processing, and radio frequency (RF) systems. In addition to contributions from industry and academia, the scientific advances in these fields have also been driven in part by challenges organized by the broader community, leveraging large-scale datasets such as MNIST, ImageNet, and RF signal classification. Recent breakthroughs have also shown the potential of using ML to improve modeling, performance prediction, and resource provisioning in optical networks across various scales. However, despite the growing number of works, the field still lacks, to the best of our knowledge, universal benchmarks with structured datasets to guide the community in pushing forward the frontier in this research direction.
To bridge this gap, we are excited to launch the Inaugural Dataset Submission at OFC 2026, which aims to solicit datasets and to establish a collaborative platform for researchers, scholars, and practitioners from academia and industry to develop, evaluate, and showcase ML techniques applied to diverse datasets collected from real-world optical devices, platforms, and networks. Participants will have the opportunity to share their collected datasets with the community and gain first-hand experience with these datasets, including detailed hardware specifications, data collection process, and data formatting. We envision that this call for data submissions will aspire to accelerate and foster progress in the integration and use of advanced ML techniques into optical networks, and to pave the way toward smarter, more efficient, and scalable transmission systems that meet the demands of future optical systems and networks.
Topic Categories
- Optical devices and hardware performance data: Gain, noise figure, and spectral characteristics of optical amplifiers; noise and nonlinearity of fibers; laser, modulator, and detector characteristics;
- Signal and quality of transmission data: Optical channel power, OSNR and GSNR, BER, Q-factor, transmission impairments (e.g., chromatic dispersion, nonlinear effects), raw or processed signals with different modulation formats (e.g., PSK, QAM);
- Distributed acoustic, vibration, and temperature sensing data;
- Silicon photonic design data: device simulation, layout and design, experimental characterization;
- Optical network traffic, topology, wavelength and route assignment data;
- Simulators and digital twins for data generation and comparison with measurement data;
- Other types of datasets that might be of interest to the OFC community.
- Dataset Description: Submissions must include a clear and detailed description of the dataset’s contents and size, how the dataset is collected (e.g., experimental environment, setup, data collection process), how the dataset is organized (e.g., data structure and format used), and how the dataset should be used ethically and responsibly.
- Impact and Applications: Authors of each dataset submission should discuss the scientific research and/or practical applications that can be enabled by the dataset.
- Availability and Accessibility: Datasets must be available and accessible to the general public without requiring direct or personal communication with the creator(s) of the dataset. Any necessary supporting code (e.g., for loading the dataset in a specific format for post-processing) should be open-source, clearly documented, and accompanied by a README file.
- Example Use Cases and Results: Authors are encouraged, but not required, to present studies and results of example use cases and applications of the dataset.
If a dataset has been presented and used in previously published works, please also include in your submission a justification on how the dataset meets the criteria listed above.
Submission Deadline
18 November 2025, 12:00 Noon Eastern Time (UTC-05:00). This inaugural Dataset Submission will not be associated with a paper to be included in the OFC proceedings. Visit the Submit a Paper tab to submit the paper.All accepted datasets will be hosted in a dedicated repository with comprehensive documentation, and award(s) will be presented for the best dataset(s) with recognition during a dedicated session. In addition, the top dataset(s) will be considered for invitation to an IEEE/Optica JOCN specific issue. As this marks the inaugural Dataset Submission at OFC, accepted datasets will not be associated with a formal paper in the OFC proceedings.
Please contact Tingjun Chen, tingjun.chen@duke.edu, Dan Kilper, dan.kilper@tcd.ie, and Camille Delezoide, camille.delezoide@nokia-bell-labs.com, if there are any questions