Universal discrete-time reservoir computers with stochastic inputs and linear readouts using non-homogeneous state-affine systems

Item Type	Journal paper
Abstract	A new class of non-homogeneous state-affine systems is introduced for use in reservoir computing. Sufficient conditions are identified that guarantee first, that the associated reservoir computers with linear readouts are causal, time-invariant, and satisfy the fading memory property and second, that a subset of this class is universal in the category of fading memory filters with stochastic almost surely uniformly bounded inputs. This means that any discrete-time filter that satisfies the fading memory property with random inputs of that type can be uniformly approximated by elements in the non-homogeneous state-affine family.
Authors	Grigoryeva, Lyudmila & Ortega, Juan-Pablo
Journal or Publication Title	Journal of Machine Learning Research
Language	English
Subjects	computer science
HSG Classification	contribution to scientific community
HSG Profile Area	SEPS - Quantitative Economic Methods
Refereed	Yes
Date	2018
Volume	19
Page Range	1-40
Depositing User	Prof. Ph.D Juan-Pablo Ortega Lahuerta
Date Deposited	10 Nov 2019 09:48
Last Modified	20 Jul 2022 17:40
URI:	https://www.alexandria.unisg.ch/publications/258282

Download

Text
GO_JMLR.pdf
Download (619kB)

Citation

Grigoryeva, Lyudmila & Ortega, Juan-Pablo (2018) Universal discrete-time reservoir computers with stochastic inputs and linear readouts using non-homogeneous state-affine systems. Journal of Machine Learning Research, 19 1-40.

Statistics

https://www.alexandria.unisg.ch/id/eprint/258282

Edit item