| Authors: | M. Freiberger, A. Katumba, P. Bienstman, J. Dambre | | Title: | Training Passive Photonic Reservoirs with Integrated Optical Readout | | Format: | International Journal | | Publication date: | 7/2019 | | Journal/Conference/Book: | IEEE Transactions on Neural Networks and Learning Systems
| | Editor/Publisher: | IEEE, | | Volume(Issue): | 30(7) p.1943-1953 | | DOI: | 10.1109/tnnls.2018.2874571 | | Citations: | 30 (Dimensions.ai - last update: 17/11/2024)
18 (OpenCitations - last update: 27/6/2024)
Look up on Google Scholar
|
Abstract
As Moore's law is coming to an end, neuromorphic approaches to computing are on the rise. One of them, passive photonic reservoir computing, is a strong candidate for energy efficient computing at high bitrates (> 10 Gbps). Currently though, both benefits are limited by the necessity to perform training and readout operations in the electrical domain. Thus, efforts are currently made in the photonic community to design an integrated optical readout, which allows to perform all operations in the optical domain. In addition to the technological challenge of designing such a readout, new algorithms have to be found in order to train it in the optical domain. Foremost, suitable algorithms need to be able to deal with the fact that the actual onchip reservoir states are not directly observable. In this work, we
investigate the possible options for such a training algorithm and propose a solution in which the complex states of the reservoir can be observed by appropriately setting the readout weights, while iterating over a predefined input sequence. We perform numerical simulations in order to compare our method with an ideal baseline requiring full observability as well as a black-box optimization method (CMA-ES).
Related Research Topics
Related Projects
|
|
|
|
|
Citations (OpenCitations)
|
|
