CALCULATION DEVICE BASED ON HIGH DENSITY OPTICAL NEURAL NETWORKS
The device builds a high-density optical artificial neural network for simultaneous processing of input signals. The invention exploits suitably diffused laser beams to generate random fields capable of replicating the functioning of classical artificial neural networks, improving their performance in terms of efficiency and processing speed.
The technology is based on the processes of coherent light scattering and interference. Intensity patterns produced by an irregular diffuser illuminated by multiple modulated laser beams are appropriately encoded to solve multiple complex Boolean functions simultaneously, avoiding the use of time- and energy-inefficient learning algorithms. Unlike classical artificial neural networks, the device enables the processing of large amounts of ‘information, without the use of high-performance processors and memories. This technology opens new perspectives for the design of systems for “edge computing,” image processing, pattern recognition, and parallel computing.
Declared TRL: 4
- Image recognition and processing;
- Edge computing;
- Digital processing;
- Data encryption.
- Allows fast processing of digital information;
- Does not require inefficient learning algorithms;
- Does not require the use of microprocessors or high-performance memories;
- Implements parallel computing.