Publications by authors named "Lorenzo De Marinis"
Article Synopsis
- This paper presents optics-informed Neural Networks (NNs) designed to enhance the performance of End-to-End deep learning models for optical transmission links using Intensity Modulation with Direct Detection (IM/DD).
- The NNs utilize building blocks derived from the response of photonic devices and incorporate a new activation function called the Photonic Sigmoid, based on a semiconductor nonlinear optical module.
- Experimental results show that these optics-informed models outperform traditional ReLU-based configurations, achieving improved noise and chromatic dispersion compensation, with performance reaching below the Bit Error Rate (BER) limit for fiber lengths up to 42 km at 48 Gb/s.
In this paper we present CHARLES (C++ pHotonic Aware neuRaL nEtworkS), a C++ library aimed at providing a flexible tool to simulate the behavior of Photonic-Aware Neural Network (PANN). PANNs are neural network architectures aware of the constraints due to the underlying photonic hardware, mostly in terms of low equivalent precision of the computations. For this reason, CHARLES exploits fixed-point computations for inference, while it supports both floating-point and fixed-point numerical formats for training.
View Article and Find Full Text PDF