Reservoir Computing with Delayed Input for Fast and Easy Optimisation

Abstract

Reservoir computing is a machine learning method that solves tasks using the response of a dynamical system to a certain input. As the training scheme only involves optimising the weights of the responses of the dynamical system, this method is particularly suited for hardware implementation. Furthermore, the inherent memory of dynamical systems which are suitable for use as reservoirs mean that this method has the potential to perform well on time series prediction tasks, as well as other tasks with time dependence. However, reservoir computing still requires extensive task-dependent parameter optimisation in order to achieve good performance. We demonstrate that by including a time-delayed version of the input for various time series prediction tasks, good performance can be achieved with an unoptimised reservoir. Furthermore, we show that by including the appropriate time-delayed input, one unaltered reservoir can perform well on six different time series prediction tasks at a very low computational expense. Our approach is of particular relevance to hardware implemented reservoirs, as one does not necessarily have access to pertinent optimisation parameters in physical systems but the inclusion of an additional input is generally possible.

Reducing reservoir computer hyperparameter dependence by external timescale tailoring

“…We consider the influence of one additional time-delayed input implemented as in [49,50]. The input function given by equation ( 3) is modified to…”

“…For the internal delay scan N was increased in steps of S starting at N = 1. This was done to avoid resonances between the delay N and clock cycle T, which are known to be detrimental [33,49].…”

Time Shifts to Reduce the Size of Reservoir Computers

“…One way to do this is to oversample the reservoir; this has been shown to improve performance for small oversampling factors such that the number of "effective nodes" is increased by a factor of 2-3 [39,40,42]. Jaurigue et al [38] showed that augmenting the reservoir matrix with a delayed version of the input signal can also improve the quality of time series prediction.…”

Multiplexed random-access optical memory in warm cesium vapor

“…Quantum memories are considered to be a main component for the realization of many future second generation quantum technologies. Their potential use ranges from synchronizing inputs into various types of quantum systems 1 to re-configurable optical reservoir computing 2 . They enable on-demand operation of otherwise probabilistic single-photon sources and quantum gates 3 , which will significantly enhance their rate of operation 4 .…”