Abstract

This article presents the experimental demonstration of a true-time delay line optical reservoir computer (ORC) using an incoherent light chaos oscillator. In contrast to other benchtop ORC systems, no external fiber spool was employed, enabling a characteristic delay of 28.4 ns, one of the fastest reported optical reservoirs to date. Comparable error metrics were obtained for standard benchmark tasks despite the reduced time scale. Practical experimental techniques, namely preamble functions and fading memory capacity measures, are introduced in this article. A mathematical model of optimized virtual nodes for the best performance of the RC was established. The fast ORC was also applied for two real world applications: respiratory motion prediction used in radiotherapy and perovskite compound property prediction used in photovoltaic material discovery. The respiratory motion prediction was compared with long short-term memory (LSTM) machine learning algorithms, the former attaining compatible results with orders of magnitude faster training speed. The ORC results for the perovskite compound classification task were compared with random forest approach, where the former demonstrated slightly better computation predication but again with much faster computing speed.

PDF Article

References

You do not have subscription access to this journal. Citation lists with outbound citation links are available to subscribers only. You may subscribe either as an OSA member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access OSA Member Subscription

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an OSA member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access OSA Member Subscription