Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware
Programming recurrent spiking neural networks (RSNNs) to robustly perform multi-timescale computation remains a difficult challenge. To address this, we describe a single-shot weight learning scheme to embed robust multi-timescale dynamics into attractor-based RSNNs, by exploiting the properties of...
Saved in:
Main Authors: | , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IOP Publishing
2025-01-01
|
Series: | Neuromorphic Computing and Engineering |
Subjects: | |
Online Access: | https://doi.org/10.1088/2634-4386/ada851 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
_version_ | 1825206181083217920 |
---|---|
author | Madison Cotteret Hugh Greatorex Alpha Renner Junren Chen Emre Neftci Huaqiang Wu Giacomo Indiveri Martin Ziegler Elisabetta Chicca |
author_facet | Madison Cotteret Hugh Greatorex Alpha Renner Junren Chen Emre Neftci Huaqiang Wu Giacomo Indiveri Martin Ziegler Elisabetta Chicca |
author_sort | Madison Cotteret |
collection | DOAJ |
description | Programming recurrent spiking neural networks (RSNNs) to robustly perform multi-timescale computation remains a difficult challenge. To address this, we describe a single-shot weight learning scheme to embed robust multi-timescale dynamics into attractor-based RSNNs, by exploiting the properties of high-dimensional distributed representations. We embed finite state machines into the RSNN dynamics by superimposing a symmetric autoassociative weight matrix and asymmetric transition terms, which are each formed by the vector binding of an input and heteroassociative outer-products between states. Our approach is validated through simulations with highly nonideal weights; an experimental closed-loop memristive hardware setup; and on Loihi 2, where it scales seamlessly to large state machines. This work introduces a scalable approach to embed robust symbolic computation through recurrent dynamics into neuromorphic hardware, without requiring parameter fine-tuning or significant platform-specific optimisation. Moreover, it demonstrates that distributed symbolic representations serve as a highly capable representation-invariant language for cognitive algorithms in neuromorphic hardware. |
format | Article |
id | doaj-art-b723736608b641d78c9247bd11a5095c |
institution | Kabale University |
issn | 2634-4386 |
language | English |
publishDate | 2025-01-01 |
publisher | IOP Publishing |
record_format | Article |
series | Neuromorphic Computing and Engineering |
spelling | doaj-art-b723736608b641d78c9247bd11a5095c2025-02-07T12:46:01ZengIOP PublishingNeuromorphic Computing and Engineering2634-43862025-01-015101400810.1088/2634-4386/ada851Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardwareMadison Cotteret0https://orcid.org/0000-0002-4891-4835Hugh Greatorex1https://orcid.org/0000-0002-3716-3992Alpha Renner2https://orcid.org/0000-0002-0724-4169Junren Chen3https://orcid.org/0009-0000-9851-0903Emre Neftci4Huaqiang Wu5https://orcid.org/0000-0001-8359-7997Giacomo Indiveri6https://orcid.org/0000-0002-7109-1689Martin Ziegler7Elisabetta Chicca8https://orcid.org/0000-0002-5518-8990Bio-Inspired Circuits and Systems (BICS) Lab, Zernike Institute for Advanced Materials University of Groningen , Groningen, The Netherlands; Groningen Cognitive Systems and Materials Center (CogniGron) University of Groningen , Groningen, The Netherlands; Micro- and Nanoelectronic Systems (MNES), Technische Universität Ilmenau , Ilmenau, GermanyBio-Inspired Circuits and Systems (BICS) Lab, Zernike Institute for Advanced Materials University of Groningen , Groningen, The Netherlands; Groningen Cognitive Systems and Materials Center (CogniGron) University of Groningen , Groningen, The NetherlandsForschungszentrum Jülich , Jülich, GermanyInstitute of Neuroinformatics, University of Zurich and ETH Zurich , Zurich, SwitzerlandForschungszentrum Jülich , Jülich, GermanySchool of Integrated Circuits, Tsinghua University , Beijing, People’s Republic of ChinaInstitute of Neuroinformatics, University of Zurich and ETH Zurich , Zurich, SwitzerlandEnergy Materials and Devices, Department of Materials Science, Kiel University , Kiel, GermanyBio-Inspired Circuits and Systems (BICS) Lab, Zernike Institute for Advanced Materials University of Groningen , Groningen, The Netherlands; Groningen Cognitive Systems and Materials Center (CogniGron) University of Groningen , Groningen, The NetherlandsProgramming recurrent spiking neural networks (RSNNs) to robustly perform multi-timescale computation remains a difficult challenge. To address this, we describe a single-shot weight learning scheme to embed robust multi-timescale dynamics into attractor-based RSNNs, by exploiting the properties of high-dimensional distributed representations. We embed finite state machines into the RSNN dynamics by superimposing a symmetric autoassociative weight matrix and asymmetric transition terms, which are each formed by the vector binding of an input and heteroassociative outer-products between states. Our approach is validated through simulations with highly nonideal weights; an experimental closed-loop memristive hardware setup; and on Loihi 2, where it scales seamlessly to large state machines. This work introduces a scalable approach to embed robust symbolic computation through recurrent dynamics into neuromorphic hardware, without requiring parameter fine-tuning or significant platform-specific optimisation. Moreover, it demonstrates that distributed symbolic representations serve as a highly capable representation-invariant language for cognitive algorithms in neuromorphic hardware.https://doi.org/10.1088/2634-4386/ada851attractor networksdistributed computationvector symbolic architecturesneuromorphic hardware abstractionshyperdimensional computinghardware interoperability |
spellingShingle | Madison Cotteret Hugh Greatorex Alpha Renner Junren Chen Emre Neftci Huaqiang Wu Giacomo Indiveri Martin Ziegler Elisabetta Chicca Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware Neuromorphic Computing and Engineering attractor networks distributed computation vector symbolic architectures neuromorphic hardware abstractions hyperdimensional computing hardware interoperability |
title | Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware |
title_full | Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware |
title_fullStr | Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware |
title_full_unstemmed | Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware |
title_short | Distributed representations enable robust multi-timescale symbolic computation in neuromorphic hardware |
title_sort | distributed representations enable robust multi timescale symbolic computation in neuromorphic hardware |
topic | attractor networks distributed computation vector symbolic architectures neuromorphic hardware abstractions hyperdimensional computing hardware interoperability |
url | https://doi.org/10.1088/2634-4386/ada851 |
work_keys_str_mv | AT madisoncotteret distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT hughgreatorex distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT alpharenner distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT junrenchen distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT emreneftci distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT huaqiangwu distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT giacomoindiveri distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT martinziegler distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware AT elisabettachicca distributedrepresentationsenablerobustmultitimescalesymboliccomputationinneuromorphichardware |