Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark

Anton Akusok; Leonardo Espinosa-Leal; Kaj-Mikael Björk; Amaury Lendasse

doi:10.1007/978-3-031-55056-0_1

Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark

Anton Akusok^*
, Leonardo Espinosa-Leal
, Kaj-Mikael Björk
, Amaury Lendasse

^*Corresponding author for this work

Not published at VTT

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

Abstract

This work presents a distributed limited-memory algorithm for Extreme Learning Machines (ELM) with training data stored in Spark. The method runs batch matrix computations to obtain the direct non-iterative solution of ELM, reads data only once, and uses the least network bandwidth, making it very computationally efficient. This is achieved by extensive use of lazy evaluation and generators in Spark, deliberately avoiding operations that may lead to data caching. The method scales to virtually infinite datasets and any number of ELM neurons, with runtime being the only restricting factor.

Original language	English
Title of host publication	Proceedings of ELM 2022
Subtitle of host publication	Theory, Algorithms and Applications
Editors	Kaj-Mikael Björk
Place of Publication	Cham
Publisher	Springer
Pages	1-8
ISBN (Electronic)	978-3-031-55056-0
ISBN (Print)	978-3-031-55055-3, 978-3-031-55058-4
DOIs	https://doi.org/10.1007/978-3-031-55056-0_1
Publication status	Published - 2024
MoE publication type	A4 Article in a conference publication
Event	12th International Conference on Extreme Learning Machines (ELM 2022) - Helsinki, Finland Duration: 8 Dec 2022 → 9 Dec 2022

Publication series

Series	Proceedings in Adaptation, Learning and Optimization
Volume	18
ISSN	2363-6084

Conference

Conference	12th International Conference on Extreme Learning Machines (ELM 2022)
Country/Territory	Finland
City	Helsinki
Period	8/12/22 → 9/12/22

Access to Document

10.1007/978-3-031-55056-0_1

Cite this

@inproceedings{c78b17facad64c92b45db5baa8d3a571,

title = "Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark",

abstract = "This work presents a distributed limited-memory algorithm for Extreme Learning Machines (ELM) with training data stored in Spark. The method runs batch matrix computations to obtain the direct non-iterative solution of ELM, reads data only once, and uses the least network bandwidth, making it very computationally efficient. This is achieved by extensive use of lazy evaluation and generators in Spark, deliberately avoiding operations that may lead to data caching. The method scales to virtually infinite datasets and any number of ELM neurons, with runtime being the only restricting factor.",

author = "Anton Akusok and Leonardo Espinosa-Leal and Kaj-Mikael Bj{\"o}rk and Amaury Lendasse",

year = "2024",

doi = "10.1007/978-3-031-55056-0\_1",

language = "English",

isbn = "978-3-031-55055-3",

series = "Proceedings in Adaptation, Learning and Optimization",

publisher = "Springer",

pages = "1--8",

editor = "Kaj-Mikael Bj{\"o}rk",

booktitle = "Proceedings of ELM 2022",

address = "Germany",

note = "12th International Conference on Extreme Learning Machines (ELM 2022) ; Conference date: 08-12-2022 Through 09-12-2022",

}

Akusok, A, Espinosa-Leal, L, Björk, K-M & Lendasse, A 2024, Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark. in K-M Björk (ed.), Proceedings of ELM 2022: Theory, Algorithms and Applications. Springer, Cham, Proceedings in Adaptation, Learning and Optimization, vol. 18, pp. 1-8, 12th International Conference on Extreme Learning Machines (ELM 2022), Helsinki, Finland, 8/12/22. https://doi.org/10.1007/978-3-031-55056-0_1

Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark. / Akusok, Anton; Espinosa-Leal, Leonardo; Björk, Kaj-Mikael et al.
Proceedings of ELM 2022: Theory, Algorithms and Applications. ed. / Kaj-Mikael Björk. Cham: Springer, 2024. p. 1-8 (Proceedings in Adaptation, Learning and Optimization, Vol. 18).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Distributed Memory-Efficient Algorithm for Extreme Learning Machines Based on Spark

AU - Akusok, Anton

AU - Espinosa-Leal, Leonardo

AU - Björk, Kaj-Mikael

AU - Lendasse, Amaury

PY - 2024

Y1 - 2024

N2 - This work presents a distributed limited-memory algorithm for Extreme Learning Machines (ELM) with training data stored in Spark. The method runs batch matrix computations to obtain the direct non-iterative solution of ELM, reads data only once, and uses the least network bandwidth, making it very computationally efficient. This is achieved by extensive use of lazy evaluation and generators in Spark, deliberately avoiding operations that may lead to data caching. The method scales to virtually infinite datasets and any number of ELM neurons, with runtime being the only restricting factor.

AB - This work presents a distributed limited-memory algorithm for Extreme Learning Machines (ELM) with training data stored in Spark. The method runs batch matrix computations to obtain the direct non-iterative solution of ELM, reads data only once, and uses the least network bandwidth, making it very computationally efficient. This is achieved by extensive use of lazy evaluation and generators in Spark, deliberately avoiding operations that may lead to data caching. The method scales to virtually infinite datasets and any number of ELM neurons, with runtime being the only restricting factor.

UR - https://github.com/akusok/pyspark-elm

U2 - 10.1007/978-3-031-55056-0_1

DO - 10.1007/978-3-031-55056-0_1

M3 - Conference article in proceedings

SN - 978-3-031-55055-3

SN - 978-3-031-55058-4

T3 - Proceedings in Adaptation, Learning and Optimization

SP - 1

EP - 8

BT - Proceedings of ELM 2022

A2 - Björk, Kaj-Mikael

PB - Springer

CY - Cham

T2 - 12th International Conference on Extreme Learning Machines (ELM 2022)

Y2 - 8 December 2022 through 9 December 2022

ER -