Optical signal processing for CNN´s

The authors point out that signal processing in cellular neural networks (CNNs) consists of multiplying and adding of signals, of calculating correlations, and of shaping signals with a nonlinear function. These operations are transformed into optical processing steps by using spatial light modulators (SLMs) and lenses as basic optical elements. The SLM performs multiplications massively in parallel and with the speed of light, whereas lenses realize a Fourier transform. A special system with two SLMs and two lenses is able to represent an array of CNNs without severe limitations on the interconnections between neurons. An example of optical pattern recognition is presented