Evolino for recurrent support vector machines

Jürgen Schmidhuber; Matteo Gagliolo; Daan Wierstra; Faustino Gomez

Evolino for recurrent support vector machines

Jürgen Schmidhuber, Matteo Gagliolo, Daan Wierstra, Faustino Gomez

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We introduce a new class of recurrent, truly sequential SVM-like devices with internal adaptive states, trained by a novel method called EVOlution of systems with KErnel-based outputs (Evoke), an instance of the recent Evolino class of methods. Evoke evolves recurrent networks to detect and represent temporal dependencies while using SVM to produce precise outputs. Evoke is the first SVM-based mechanism learning to classify a context-sensitive language. It also outperforms recent state-of-the-art gradient-based recurrent neural networks (RNNs) on various time series prediction tasks.

Original language	English (US)
Title of host publication	ESANN 2006 Proceedings - European Symposium on Artificial Neural Networks
Publisher	d-side publication
Pages	593-598
Number of pages	6
ISBN (Print)	2930307064
State	Published - Jan 1 2006
Externally published	Yes

Cite this

@inproceedings{0fd5dccf19424eb1b75ebab100569339,

title = "Evolino for recurrent support vector machines",

abstract = "We introduce a new class of recurrent, truly sequential SVM-like devices with internal adaptive states, trained by a novel method called EVOlution of systems with KErnel-based outputs (Evoke), an instance of the recent Evolino class of methods. Evoke evolves recurrent networks to detect and represent temporal dependencies while using SVM to produce precise outputs. Evoke is the first SVM-based mechanism learning to classify a context-sensitive language. It also outperforms recent state-of-the-art gradient-based recurrent neural networks (RNNs) on various time series prediction tasks.",

author = "J{\"u}rgen Schmidhuber and Matteo Gagliolo and Daan Wierstra and Faustino Gomez",

note = "Generated from Scopus record by KAUST IRTS on 2022-09-14",

year = "2006",

month = jan,

day = "1",

language = "English (US)",

isbn = "2930307064",

pages = "593--598",

booktitle = "ESANN 2006 Proceedings - European Symposium on Artificial Neural Networks",

publisher = "d-side publication",

}

TY - GEN

T1 - Evolino for recurrent support vector machines

AU - Schmidhuber, Jürgen

AU - Gagliolo, Matteo

AU - Wierstra, Daan

AU - Gomez, Faustino

N1 - Generated from Scopus record by KAUST IRTS on 2022-09-14

PY - 2006/1/1

Y1 - 2006/1/1

N2 - We introduce a new class of recurrent, truly sequential SVM-like devices with internal adaptive states, trained by a novel method called EVOlution of systems with KErnel-based outputs (Evoke), an instance of the recent Evolino class of methods. Evoke evolves recurrent networks to detect and represent temporal dependencies while using SVM to produce precise outputs. Evoke is the first SVM-based mechanism learning to classify a context-sensitive language. It also outperforms recent state-of-the-art gradient-based recurrent neural networks (RNNs) on various time series prediction tasks.

AB - We introduce a new class of recurrent, truly sequential SVM-like devices with internal adaptive states, trained by a novel method called EVOlution of systems with KErnel-based outputs (Evoke), an instance of the recent Evolino class of methods. Evoke evolves recurrent networks to detect and represent temporal dependencies while using SVM to produce precise outputs. Evoke is the first SVM-based mechanism learning to classify a context-sensitive language. It also outperforms recent state-of-the-art gradient-based recurrent neural networks (RNNs) on various time series prediction tasks.

UR - http://www.scopus.com/inward/record.url?scp=34250029157&partnerID=8YFLogxK

M3 - Conference contribution

SN - 2930307064

SP - 593

EP - 598

BT - ESANN 2006 Proceedings - European Symposium on Artificial Neural Networks

PB - d-side publication

ER -

Evolino for recurrent support vector machines

Abstract

Other files and links

Fingerprint

Cite this