Abstract

In this work, we propose and numerically investigate a scheme for reservoir computing (RC) based on two parallel reservoirs under identical electrical message injection, in which two semiconductor lasers (SLs) under optical feedback are utilized as two parallel reservoirs. For simplifying the system, only one mask signal is employed in this scheme. After multiplying with input data, the masked information is injected into two parallel reservoir lasers (SL1 and SL2). The virtual node states can be obtained from the temporal outputs of two SLs. RC can be accomplished by three ways, namely RC1/RC2 (the virtual node states originating from SL1/SL2 are used for training and testing) and RCM (the merged virtual node states originating from two SLs are used for training and testing). Via chaotic time series prediction task and memory capacity (MC) test, the performance of the RC system is simulated and assessed. The results show that, for a given data processing rate, better prediction performance and higher MC can be realized by RCM through setting suitable mismatched parameters between the two SLs. Under satisfying the requirement for achieving a good performance, the highest data processing rate can be doubled for RCM with respect to that for RC1/RC2.

Highlights

  • Reservoir computing (RC), originally introduced as an echo state network and a liquid state machine [1], [2], is a novel information processing technique emerged in the field of machine learning and has evolved into a popular algorithm for processing complex tasks in recent years [3]

  • The prediction performance is quantitatively measured by the normalized mean square error (NMSE), which is expressed as: NMSE = 1 L (y (n) − y(n))2/var (y) n=1 where L is the total number of tested data, y is the output of RC, yis the target, and var denotes the variance

  • We proposed and numerically demonstrated an RC scheme based on two parallel reservoirs under identical electrical message injection, where the two reservoirs are composed of two semiconductor lasers (SLs) (SL1 and SL2) subjected to optical feedback

Read more

Summary

Introduction

Reservoir computing (RC), originally introduced as an echo state network and a liquid state machine [1], [2], is a novel information processing technique emerged in the field of machine learning and has evolved into a popular algorithm for processing complex tasks in recent years [3]. To further increase data processing rate, one can construct an RC system based on ensembles of time-delay reservoirs as introduced in [33], [34], where multiple reservoirs are coupled or uncoupled each other. We noticed that in relevant reports including the above two works [35], [36], the input data is firstly multiplied by different masks to form different information, and the information is optically injected into multiple reservoirs, respectively. [35], two SLs subjected to optical feedback are taken as two parallel reservoirs in such a RC system, and the information masked by identical mask signal is injected into two parallel reservoirs through modulating the pump current of two SLs. Different from that reported in Ref. Via chaotic series prediction task and memory capacity (MC) test, the performance of the RC system based on two parallel reservoirs are evaluated and compared with that based on a single reservoir

System Model
Santa Fe Time Series Prediction
Memory Capacity
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call