Next generation reservoir computing

Gauthier, Daniel J.; Bollt, Erik M.; Griffith, Aaron; Barbosa, Wendson A. S.

doi:10.1038/s41467-021-25801-2

Cited by 296 publications

(232 citation statements)

References 35 publications

Supporting

Mentioning

180

Contrasting

Order By: Relevance

“…In both cases there is an improvement in the performance with the correct choice of the delayed input. It is has been demonstrated that the Lorenz x one step ahead prediction requires only the very recent history of the x-variable time series [13], and we find the optimal input delay of d = 1 to be consistent with this prior knowledge. For the Lorenz z cross-prediction task, on the other hand, there is a strong dependence on the history of the Lorenz x variable.…”

Section: Resultssupporting

confidence: 68%

See 1 more Smart Citation

Reservoir Computing with Delayed Input for Fast and Easy Optimisation

Jaurigue

Robertson

Wolters

et al. 2021

Entropy

View full text Add to dashboard Cite

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.

show abstract

Section: Resultssupporting

confidence: 68%

“…In a recent paper [ 13 ], the authors aim to eliminate the issue of hyperparameter optimisation altogether by removing the reservoir. Their approach essentially takes the well-known nonlinear vector autoregression (NVAR) method, uses a less parsimonious approach to filling the feature vector, and adds Tikhonov regularisation.…”

Section: Introductionmentioning

confidence: 99%

Reservoir Computing with Delayed Input for Fast and Easy Optimisation

Jaurigue

Robertson

Wolters

et al. 2021

Entropy

View full text Add to dashboard Cite

show abstract

“…3a 126 with G 2 = 0) is reasonable. Moreover, this finding supports the general observation that only the very recent history of the x-variable time series [13], and we find the optimal parameters and was also demonstrated for a time-continuous system, indicating that 180 our approach is applicable to a wide range of reservoirs. for a reservoir to yield good performance on a range of tasks by only tuning the delayed 202 input parameters, remain to be determined.…”

supporting

confidence: 86%

“…In a recent paper [13], the authors aim to eliminate the issue of hyperparameter 31 optimisation altogether by removing the reservoir. Their approach essentially takes the 32 well known nonlinear vector autoregression (NVAR) method, uses a less parsimonious 33 approach to filling the feature vector and adds Tikhonov regularisation.…”

mentioning

confidence: 99%

“…Their approach essentially takes the 32 well known nonlinear vector autoregression (NVAR) method, uses a less parsimonious 33 approach to filling the feature vector and adds Tikhonov regularisation. However, 34 the method of [13] trades the optimisation of the reservoir hyperparameters for the 35 optimisation of the feature vector elements and it can not be asserted that the latter is 36 generally less costly. Furthermore, one of the main factors driving research into reservoir 37 have not been used to optimise the performance of reservoir computers.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Reservoir Computing with Delayed Input for Fast and Easy Optimization

Jaurigue¹,

Robertson²,

Wolters³

et al. 2021

Preprint

View full text Add to dashboard Cite

Reservoir computing is a machine learning method that uses the response of a dynamical system to a certain input in order to solve a task. 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.

show abstract

Circuit‐Level Memory Technologies and Applications based on 2D Materials

Liu

Yang

et al. 2022

Advanced Materials

View full text Add to dashboard Cite

random-access memory (DRAMs), static random-access memory (SRAMs), [1] and flash memory, [2] multiple transistors and capacitors are employed to constitute a single functional cell. These functional cells are typically integrated to form a memory array for advanced functionality, miniaturized footprints, low power consumption, and reduced latency. Traditional silicon-based complementary metal-oxidesemiconductor (CMOS) technology has been widely used to build these memory arrays. However, the performance improvement of CMOS-based memory relies on the advance of nanofabrication technology to scale down the feature size of transistors, which is approaching the physical limit. [3] To overcome this bottleneck, emerging memory technologies are brought up as promising supplements to conventional memory devices. [4] These novel types of memory include resistive random-access memory (RRAM), [5,6] ferroelectric RAM (FeRAM), [7] optoelectronic RRAM (ORRAM), [8,9] phase-change memory (PCM), [10] and spin-transfer torque magnetoresistive RAM (STT-MRAM). [11,12] Furthermore, emerging memory devices allow non-von-Neumann architectures that pave the way toward high-performance in-memory computing technology. [13,14] For example, memristive crossbar arrays composed of 1-transistor-1-resistor (1T1R) unit cells can enable in-memory computing and are the most cost-efficient thanks to their two-terminal structure. [15] Advanced 3D monolithic stacking integration also emerges as a promising approach to Memory technologies and applications implemented fully or partially using emerging 2D materials have attracted increasing interest in the research community in recent years. Their unique characteristics provide new possibilities for highly integrated circuits with superior performances and low power consumption, as well as special functionalities. Here, an overview of progress in 2D-material-based memory technologies and applications on the circuit level is presented. In the material growth and fabrication aspects, the advantages and disadvantages of various methods for producing large-scale 2D memory devices are discussed. Reports on 2D-material-based integrated memory circuits, from conventional dynamic random-access memory, static random-access memory, and flash memory arrays, to emerging memristive crossbar structures, all the way to 3D monolithic stacking architecture, are systematically reviewed. Comparisons between experimental implementations and theoretical estimations for different integration architectures are given in terms of the critical parameters in 2D memory devices. Attempts to use 2D memory arrays for in-memory computing applications, mostly on logic-in-memory and neuromorphic computing, are summarized here. Finally, challenges that impede the large-scale applications of 2D-material-based memory are reviewed, and perspectives on possible approaches toward a more reliable system-level fabrication are also given, hopefully shedding some light on future research.

show abstract

Next generation reservoir computing

Cited by 296 publications

References 35 publications

Reservoir Computing with Delayed Input for Fast and Easy Optimisation

Reservoir Computing with Delayed Input for Fast and Easy Optimisation

Reservoir Computing with Delayed Input for Fast and Easy Optimization

Circuit‐Level Memory Technologies and Applications based on 2D Materials

Contact Info

Product

Resources

About