Neural network based optical module for transmission-reception
Introduction
The photonic integrated device is effective to transmit and recover optical signal. It integrates an optical network able to correct both linear and nonlinear distorsions, avoiding electro-optic conversion steps and working transparently on the optical link.

Technical features
The present invention refers to a photonic integrated module acting both in transmission and in reception mode and it includes a neural network. The device is based on a series of interferometers coupled with nonlinear elements and it can be either directly integrated onto an optical transponder or acting as a transparent block on an optical transmission line. The known neural networks are unable to process the signal while keeping it in the optical field. It does not require any electrostatic conversion to process the signal. The device is extremely compact and processes signals at frequencies that are difficult to reach by electronic systems; moreover, it works in the optical domain and is able to recover the signal by acting on both linear dispersion as well as optical link nonlinearities. Its use greatly relaxes the need of custom microprocessors and, for some applications, it can work in a fully passive mode. The optical link does not require any modification, as the device act on the optical domain and is fully transparent to the network.
Possible Applications
- Optical communications;
- Signal transmission and recovery;
- Optical networks;
- Optical transponders;
- Internet of things.
Advantages
- Transparent installation even on legacy hardwares;
- Works on the optical domain without E/O conversions;
- Integrated device;
- Small footprint (compatible with the existent form factors);
- Low cost;
- Low power consumption.