Parallel dual optical feedback semiconductor laser reservoir computing system

Abstract

The invention discloses a parallel dual optical feedback semiconductor laser reservoir computing system which comprises an input layer, a reservoir layer and an output layer. The input layer includes a driving laser, an arbitrary waveform generator, a modulator, and a coupler I and is used for dividing an input signal into two parts to be input into the reservoir layer. The reservoir layer consists of two response lasers with dual optical feedback loops and is used for generating a rich nonlinear dynamic response. Each optical feedback loop is composed of a circulator, a delay fiber, and an adjustable attenuator. Each response laser has a feedback loop and two couplers and is used for receiving the input signal and outputting a portion of light to the output layer. The output layer includes a pair of photoelectric detectors and a dual-channel digitizer to obtain the response status of the reservoir. The parallel dual optical feedback semiconductor laser reservoir computing system can reduce the number of virtual nodes of the feedback loop receiving the input signal and the period of a mask signal without degrading the performance, thereby reducing the requirement of the reservoir for the cache size of the arbitrary waveform generator.

Description

technical field [0001] The invention belongs to the technical field of optical storage pool computing, and in particular relates to a parallel dual optical feedback semiconductor laser storage pool computing system. Background technique [0002] As a machine learning method, reserve pool computing has broad application prospects in many scenarios with its simple training method and good performance. The semiconductor laser storage pool computing system is a hardware storage pool computing system. It realizes reserve pool calculation through optical devices, which can speed up the processing speed by taking advantage of the fast signal processing speed of light. The semiconductor laser storage pool computing system is often implemented with a single semiconductor laser plus a delay feedback loop structure. Each input data must maintain a delay time of the feedback loop, and after multiplied by the mask signal, it is serially input into the reserve pool. These data need to ...

AI Technical Summary

Problems solved by technology

These data need to be generated by an arbitrary waveform generator. When the length of the input data and the number of virtual nodes of the feedback loop (corresponding to the period of the mask signal) are large, the arbitrary waveform generator is required to generate a large amount of data at the same time. However, the arbitrary waveform generator with a large buffer is often very expensive, which makes the implementation cost of the optical storage pool computing system high and difficult to be practical

Images