Editorial: Machine-learning-assisted photonic design: from fundamental physics to advanced devices

Over the past few decades, machine learning (ML) has profoundly influenced a broad range of scientific and technological disciplines, enabling the exploration of increasingly complex problems through powerful data-driven methodologies. The convergence of machine learning with photonic design represents one of the most dynamic intersections between modern computational science and fundamental physics. At the same time, rapid advances in optical materials and device technologies have opened new frontiers in data communications, computing, imaging, sensing, and biophotonics.

As photonics evolves from conventional linear optical components toward sophisticated multi-scale platforms such as metasurfaces, nonlinear metamaterials, and integrated photonic circuits, traditional trial-and-error approaches and purely analytical design methodologies become increasingly impractical. The enormous design space of modern photonic systems, together with the highly nonlinear interactions between light and complex materials, calls for fundamentally new engineering strategies. In this context, machine learning has emerged as a transformative tool for the design, optimization, and control of advanced optical devices and materials.

Importantly, purely data-driven ML approaches often struggle to capture the underlying physical principles that govern complex optical phenomena. By contrast, physics-informed and physics-inspired machine learning frameworks integrate physical constraints and prior knowledge directly into the learning process, yielding models that are more accurate, robust, interpretable, and computationally efficient. These approaches are not only improving our understanding of light propagation and nonlinear optical dynamics but also accelerating the discovery of novel materials and enabling the inverse design of photonic structures with targeted functionalities.

Despite these significant advances, important challenges remain in fully understanding and exploiting the capabilities of machine learning for photonic design, motivating continued research at the interface of physics, computation, and engineering.

This Research Topic brings together contributions that employ machine learning methods for the design, analysis, and control of advanced optical systems. The collected works span both data-driven and physics-based approaches, demonstrating how intelligent computational methods can accelerate photonic device development while providing deeper insight into the underlying physical mechanisms.

A central theme emerging from this Research Topic is the use of machine learning to uncover and model the fundamental dynamics of light propagation in nonlinear and complex media. At the foundational level, understanding nonlinear wave dynamics is essential for the development of all-optical computing and signal-processing technologies. In this direction, Yang et al. introduce a data-driven framework based on Sparse Identification of Nonlinear Dynamics (SINDy) to numerically capture the evolution of moment quantities associated with the nonlinear Schrödinger (NLS) equation. By combining sparse regression and optimization techniques, the authors demonstrate that the method can effectively reproduce complex nonlinear dynamics, rediscover known analytical closures, and provide reduced-order descriptions even for systems lacking exact analytical formulations.

Physics-inspired neural architectures also play an increasingly important role in inverse problems arising in optical sensing and spectroscopy. In medical and biomedical imaging, retrieving optical properties from diffuse scattering media remains computationally demanding when based solely on traditional diffusion theory. Witteveen et al. address this challenge by developing a physics-inspired neural network that extracts physiological parameters from diffuse reflectance spectroscopy and hyperspectral imaging data. Trained on diffusion-theory-generated datasets, the physics-driven model achieves accurate retrieval of quantities such as blood volume fraction and scattering amplitudes while dramatically reducing computational cost relative to conventional inversion methods.

Machine learning techniques are likewise proving valuable in parameter optimization and experimental design problems. In mid-infrared spectroscopy, protein and peptide characterization requires identifying optimal wavelength ranges across broad spectral domains. Aledda et al. investigate several machine learning strategies for wavelength selection in tunable laser systems, including sparse modeling, filter-based approaches, and compression algorithms, and introduce a new method, w-CovSel. Their work demonstrates how data-driven optimization can improve the performance and efficiency of laser-based analytical instruments tailored to specific biochemical applications such as peptide analysis.

This Research Topic also highlights the growing shift toward photonics for machine learning, where optical hardware itself is used to accelerate computational tasks. While much of the current research in the field applies ML to photonic design, an emerging direction seeks to embed machine learning algorithms directly within optical systems to achieve ultrafast and energy-efficient information processing. Mahmoud et al. contribute to this direction by proposing an Optical Particle Swarm Optimization (OPSO) algorithm that employs an all-optical computational update mechanism. By exploiting the natural properties of coherent optical dynamics, such as nonlinear light-matter interactions and complex-domain computation, their approach demonstrates that optical computing can significantly enhance search-space exploration and improve dyslexia detection performance using eye-tracking data.

Taken together, the studies presented in this Research Topic indicate that the field of photonics is entering a new era characterized by the synergistic integration of physics-informed artificial intelligence, nonlinear optical science, and optical computing hardware. The field is evolving beyond computationally assisted optimization toward increasingly autonomous and intelligent photonic design paradigms, where machine learning algorithms operate under explicit physical constraints and interact directly with optical hardware platforms.

Looking forward, substantial challenges remain. Progress will require access to large, high-quality datasets, while experimental data acquisition - particularly in sensitive fields such as biophotonics - often remains costly and difficult. Furthermore, transitioning from software-based machine learning models to scalable physical optical neural networks presents major challenges associated with scalability, robustness, and signal-to-noise limitations. Nevertheless, the contributions collected in this Research Topic provide a strong foundation for addressing these obstacles and point toward a future in which machine learning and photonics become deeply integrated components of a unified technological framework for scientific discovery and advanced device engineering.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited and Reviewed by: Francesco Morichetti, Polytechnic University of Milan, Italy