Fast on-device adaptation for spiking neural networks via online-within-online meta-learning

Bleema Rosenfeld; Bipin Rajendran; Osvaldo Simeone

doi:10.1109/DSLW51110.2021.9523405

Fast on-device adaptation for spiking neural networks via online-within-online meta-learning

Bleema Rosenfeld, Bipin Rajendran, Osvaldo Simeone

Electrical and Computer Engineering

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

7 Scopus citations

Abstract

Spiking Neural Networks (SNNs) have recently gained popularity as machine learning models for on-device edge intelligence for applications such as mobile healthcare management and natural language processing due to their low power profile. In such highly personalized use cases, it is important for the model to be able to adapt to the unique features of an individual with only a minimal amount of training data. Meta-learning has been proposed as a way to train models that are geared towards quick adaptation to new tasks. The few existing meta-learning solutions for SNNs operate offline and require some form of backpropagation that is incompatible with the current neuromorphic edge-devices. In this paper, we propose an online-within-online meta-learning rule for SNNs termed OWOML-SNN, that enables lifelong learning on a stream of tasks, and relies on local, backprop-free, nested updates.

Original language	English (US)
Title of host publication	2021 IEEE Data Science and Learning Workshop, DSLW 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781665428255
DOIs	https://doi.org/10.1109/DSLW51110.2021.9523405
State	Published - Jun 5 2021
Event	2021 IEEE Data Science and Learning Workshop, DSLW 2021 - Toronto, Canada Duration: Jun 5 2021 → Jun 6 2021

Publication series

Name	2021 IEEE Data Science and Learning Workshop, DSLW 2021

Conference

Conference	2021 IEEE Data Science and Learning Workshop, DSLW 2021
Country/Territory	Canada
City	Toronto
Period	6/5/21 → 6/6/21

All Science Journal Classification (ASJC) codes

Artificial Intelligence
Information Systems
Education

Keywords

Meta-learning
Online learning
Spiking neural network

Access to Document

10.1109/DSLW51110.2021.9523405

Cite this

@inproceedings{817317f0225c48a2973d67ffb520d6c3,

title = "Fast on-device adaptation for spiking neural networks via online-within-online meta-learning",

abstract = "Spiking Neural Networks (SNNs) have recently gained popularity as machine learning models for on-device edge intelligence for applications such as mobile healthcare management and natural language processing due to their low power profile. In such highly personalized use cases, it is important for the model to be able to adapt to the unique features of an individual with only a minimal amount of training data. Meta-learning has been proposed as a way to train models that are geared towards quick adaptation to new tasks. The few existing meta-learning solutions for SNNs operate offline and require some form of backpropagation that is incompatible with the current neuromorphic edge-devices. In this paper, we propose an online-within-online meta-learning rule for SNNs termed OWOML-SNN, that enables lifelong learning on a stream of tasks, and relies on local, backprop-free, nested updates.",

keywords = "Meta-learning, Online learning, Spiking neural network",

author = "Bleema Rosenfeld and Bipin Rajendran and Osvaldo Simeone",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 2021 IEEE Data Science and Learning Workshop, DSLW 2021 ; Conference date: 05-06-2021 Through 06-06-2021",

year = "2021",

month = jun,

day = "5",

doi = "10.1109/DSLW51110.2021.9523405",

language = "English (US)",

series = "2021 IEEE Data Science and Learning Workshop, DSLW 2021",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2021 IEEE Data Science and Learning Workshop, DSLW 2021",

address = "United States",

}

Rosenfeld, B, Rajendran, B & Simeone, O 2021, Fast on-device adaptation for spiking neural networks via online-within-online meta-learning. in 2021 IEEE Data Science and Learning Workshop, DSLW 2021. 2021 IEEE Data Science and Learning Workshop, DSLW 2021, Institute of Electrical and Electronics Engineers Inc., 2021 IEEE Data Science and Learning Workshop, DSLW 2021, Toronto, Canada, 6/5/21. https://doi.org/10.1109/DSLW51110.2021.9523405

Fast on-device adaptation for spiking neural networks via online-within-online meta-learning. / Rosenfeld, Bleema; Rajendran, Bipin; Simeone, Osvaldo.
2021 IEEE Data Science and Learning Workshop, DSLW 2021. Institute of Electrical and Electronics Engineers Inc., 2021. (2021 IEEE Data Science and Learning Workshop, DSLW 2021).

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

TY - GEN

T1 - Fast on-device adaptation for spiking neural networks via online-within-online meta-learning

AU - Rosenfeld, Bleema

AU - Rajendran, Bipin

AU - Simeone, Osvaldo

PY - 2021/6/5

Y1 - 2021/6/5

N2 - Spiking Neural Networks (SNNs) have recently gained popularity as machine learning models for on-device edge intelligence for applications such as mobile healthcare management and natural language processing due to their low power profile. In such highly personalized use cases, it is important for the model to be able to adapt to the unique features of an individual with only a minimal amount of training data. Meta-learning has been proposed as a way to train models that are geared towards quick adaptation to new tasks. The few existing meta-learning solutions for SNNs operate offline and require some form of backpropagation that is incompatible with the current neuromorphic edge-devices. In this paper, we propose an online-within-online meta-learning rule for SNNs termed OWOML-SNN, that enables lifelong learning on a stream of tasks, and relies on local, backprop-free, nested updates.

AB - Spiking Neural Networks (SNNs) have recently gained popularity as machine learning models for on-device edge intelligence for applications such as mobile healthcare management and natural language processing due to their low power profile. In such highly personalized use cases, it is important for the model to be able to adapt to the unique features of an individual with only a minimal amount of training data. Meta-learning has been proposed as a way to train models that are geared towards quick adaptation to new tasks. The few existing meta-learning solutions for SNNs operate offline and require some form of backpropagation that is incompatible with the current neuromorphic edge-devices. In this paper, we propose an online-within-online meta-learning rule for SNNs termed OWOML-SNN, that enables lifelong learning on a stream of tasks, and relies on local, backprop-free, nested updates.

KW - Meta-learning

KW - Online learning

KW - Spiking neural network

UR - http://www.scopus.com/inward/record.url?scp=85115418988&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85115418988&partnerID=8YFLogxK

U2 - 10.1109/DSLW51110.2021.9523405

DO - 10.1109/DSLW51110.2021.9523405

M3 - Conference contribution

AN - SCOPUS:85115418988

T3 - 2021 IEEE Data Science and Learning Workshop, DSLW 2021

BT - 2021 IEEE Data Science and Learning Workshop, DSLW 2021

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2021 IEEE Data Science and Learning Workshop, DSLW 2021

Y2 - 5 June 2021 through 6 June 2021

ER -

Fast on-device adaptation for spiking neural networks via online-within-online meta-learning

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Keywords

Access to Document

Other files and links

Fingerprint

Cite this