Nonlinear Transformations in Telecom Devices for Optical Computing

We propose and demonstrate a photonic computing approach that leverages nonlinear transformations of optical signals in standard telecom devices. This method maps input signals to a high-dimensional feature space where linear processing can be applied, enabling efficient computation for tasks like time series forecasting and classification. The general scheme of this approach is shown in Fig. 1(a). This concept is related to the principles used in reservoir computing [1] and support vector machines [2].