Answers are grounded in this paper and cite their sources.
Beyond serving as high-capacity communication links, fiber-optic transmission systems can function as kernel platforms. Crucially, the intrinsic optical nonlinearity facilitates the projection of input data into high-dimensional feature spaces, thereby enhancing the linear separability of complex patterns required for effective classification. This capability naturally unifies intelligent computation with optical communication. This paper investigates the feasibility of achieving integrated computation and communication (ICAC) by exploiting the intrinsic and complex dynamical properties of optical fibers through three machine learning tasks: digit recognition, voice recognition, and sentiment analysis on movie review. Numerical analysis results demonstrate that by leveraging the passive fiber-optic kernel, the proposed system achieves recognition accuracies of 95.7% for MNIST digit recognition, 90.8% for voice recognition, and 84.5% for IMDB sentiment analysis. Communication recovery is also evaluated using mean square error (MSE), bit error rate (BER), and error vector magnitude (EVM). Additional fixed-kernel, scalability, and ablation analyses verify the adaptability and physical mechanism of the fiber-optic kernel.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: