Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks

The performance of echo state networks (ESNs) in temporal pattern learning tasks depends both on their memory capacity (MC) and their non-linear processing. It has been shown that linear memory capacity is maximized when ESN neurons have linear activation, and that a trade-off between non-linearity...

Full description

Saved in:
Bibliographic Details
Main Authors: Stefan Iacob, Joni Dambre
Format: Article
Language:English
Published: MDPI AG 2024-12-01
Series:Biomimetics
Subjects:
Online Access:https://www.mdpi.com/2313-7673/9/12/755
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850059192376754176
author Stefan Iacob
Joni Dambre
author_facet Stefan Iacob
Joni Dambre
author_sort Stefan Iacob
collection DOAJ
description The performance of echo state networks (ESNs) in temporal pattern learning tasks depends both on their memory capacity (MC) and their non-linear processing. It has been shown that linear memory capacity is maximized when ESN neurons have linear activation, and that a trade-off between non-linearity and linear memory capacity is required for temporal pattern learning tasks. The more recent distance-based delay networks (DDNs) have shown improved memory capacity over ESNs in several benchmark temporal pattern learning tasks. However, it has not thus far been studied whether this increased memory capacity comes at the cost of reduced non-linear processing. In this paper, we advance the hypothesis that DDNs in fact achieve a better trade-off between linear MC and non-linearity than ESNs, by showing that DDNs can have strong non-linearity with large memory spans. We tested this hypothesis using the NARMA-30 task and the bitwise delayed XOR task, two commonly used reservoir benchmark tasks that require a high degree of both non-linearity and memory.
format Article
id doaj-art-20dbcaac15954b74a95cbf384d4b971f
institution DOAJ
issn 2313-7673
language English
publishDate 2024-12-01
publisher MDPI AG
record_format Article
series Biomimetics
spelling doaj-art-20dbcaac15954b74a95cbf384d4b971f2025-08-20T02:50:56ZengMDPI AGBiomimetics2313-76732024-12-0191275510.3390/biomimetics9120755Memory–Non-Linearity Trade-Off in Distance-Based Delay NetworksStefan Iacob0Joni Dambre1IDLab-AIRO, Faculty of Engineering and Architecture, Ghent University, 9052 Ghent, BelgiumIDLab-AIRO, Faculty of Engineering and Architecture, Ghent University, 9052 Ghent, BelgiumThe performance of echo state networks (ESNs) in temporal pattern learning tasks depends both on their memory capacity (MC) and their non-linear processing. It has been shown that linear memory capacity is maximized when ESN neurons have linear activation, and that a trade-off between non-linearity and linear memory capacity is required for temporal pattern learning tasks. The more recent distance-based delay networks (DDNs) have shown improved memory capacity over ESNs in several benchmark temporal pattern learning tasks. However, it has not thus far been studied whether this increased memory capacity comes at the cost of reduced non-linear processing. In this paper, we advance the hypothesis that DDNs in fact achieve a better trade-off between linear MC and non-linearity than ESNs, by showing that DDNs can have strong non-linearity with large memory spans. We tested this hypothesis using the NARMA-30 task and the bitwise delayed XOR task, two commonly used reservoir benchmark tasks that require a high degree of both non-linearity and memory.https://www.mdpi.com/2313-7673/9/12/755information processing capacitymemory capacitymemory–non-linearity trade-offreservoir computingdistance-based delay networksecho state networks
spellingShingle Stefan Iacob
Joni Dambre
Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
Biomimetics
information processing capacity
memory capacity
memory–non-linearity trade-off
reservoir computing
distance-based delay networks
echo state networks
title Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
title_full Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
title_fullStr Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
title_full_unstemmed Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
title_short Memory–Non-Linearity Trade-Off in Distance-Based Delay Networks
title_sort memory non linearity trade off in distance based delay networks
topic information processing capacity
memory capacity
memory–non-linearity trade-off
reservoir computing
distance-based delay networks
echo state networks
url https://www.mdpi.com/2313-7673/9/12/755
work_keys_str_mv AT stefaniacob memorynonlinearitytradeoffindistancebaseddelaynetworks
AT jonidambre memorynonlinearitytradeoffindistancebaseddelaynetworks