Abstract
AbstractVarious approaches have been proposed to construct reservoir computing systems. However, the network structure and information processing capacity of these systems are often tied to their individual implementations, which typically become difficult to modify after physical setup. This limitation can hinder performance when the system is required to handle a wide spectrum of prediction tasks. To address this limitation, it is crucial to develop tunable systems that can adapt to a wide range of problem domains. This chapter presents a tunable optical computing method based on the iterative function system (IFS). The tuning capability of IFS provides adjustment of the network structure and optimizes the performance of the optical system. Numerical and experimental results show the tuning capability of the IFS reservoir computing. The relationship between tuning parameters and reservoir properties is discussed. We further investigate the impact of optical feedback on the reservoir properties and present the prediction results.
Publisher
Springer Nature Singapore
Reference27 articles.
1. Y. Lecun, Y. Bengio, G. Hinton, Deep learning. Nature 521(7553), 436–444 (2015). https://doi.org/10.1038/nature14539
2. T.W. Hughes, M. Minkov, Y. Shi, S. Fan, Training of photonic neural networks through in situ backpropagation. Optica 5(7), 864–871 (2018)
3. H. Jaeger, The “echo state” approach to analysing and training recurrent neural networks-with an erratum note, Bonn, Germany: German National Research Center for Information Technology GMD Technical Report 148(34), 13 (2001)
4. H. Jaeger, H. Haas, Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication. Science 304(5667), 78–80 (2004)
5. W. Maass, H. Markram, On the computational power of circuits of spiking neurons. J. Comput. Syst. Sci. 69(4), 593–616 (2004)