Since the 1970’s scientists and engineers have predicted the role of optical technology in computing systems, however it has always struggled to compete with electronics and the impact of Moore’s law. Recent developments in artificial intelligence, especially convolutional neural networks (CNNs) and deep learning systems, have echoed the experiments of the 1990s when optical systems were used to implement these parallel architectures. Another successful area has been optical pattern recognition or correlation, where system have been designed and implemented for undertaking complex image processing tasks. The problem with these advances has always been the faster paced progress of electronic systems, however there are now problems being proposed that even the best electronics cannot solve but are in the remit of optical technology. What is likely to succeed in these systems is a hybrid approach where optics is used to solve the high-volume parallel processing tasks and form an engine for an overall system. In this talk, the evolution of this approach will be presented and its progression into the implementation of these highly computationally intensive and parallel approaches are discussed. In particular, the work of the CMMPE team at Cambridge will show that free-space Fourier based optical systems have a huge potential to contribute to these complex problems.
June 29 @ 10:55
10:55 — 11:35 (40′)