Trainable dynamical masking for readout-free optical computing

2025/05/29 by Bogdanov, S., Manuylovich, E., Turitsyn, S. K. · 1 citation
#FOS: Physical sciences #Optics (physics.optics)

paper · doi:10.48550/arxiv.2505.23464

Abstract

Nonlinear systems, transforming an input signal into a high-dimensional output feature space, can be used for non-conventional computing. This approach, however, requires a change of system parameters during training rather than coefficients in a software program. We propose here to use available off-the-shelf high-speed optical communication devices and technologies to implement a trainable dynamical mask in addition to or even instead of the traditional readout layer for extreme learning machine-based computing. The computational potential of the proposed approach is demonstrated with both regression and time series prediction tasks.

Citations

• SOA-based reservoir computing using up-sampling
• Optical computing with supercontinuum generation in photonic crystal fibers
• The physics of optical computing
• Toward a formal theory for computing machines made out of whatever physics offers: extended version
• Programming Nonlinear Propagation for Efficient Optical Learning Machines
• Wave Physics as an Analog Recurrent Neural Network
• Recent advances in physical reservoir computing: A review
• Simulating physics with computers

Cited by

• Reservoir Computing with a single Josephson junction

Related

• Optical neuromorphic computing via temporal up-sampling and trainable encoding on a telecom device platform
• Recent Advances on Machine Learning-aided DSP for Short-reach and Long-haul Optical Communications
• Nonlinear photonic dynamical systems for unconventional computing
• Training Large-Scale Optical Neural Networks with Two-Pass Forward Propagation
• A Coherent Perceptron for All-Optical Learning