kth.sePublications KTH
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Spiking neural networks with Hebbian plasticity for unsupervised representation learning
KTH, School of Electrical Engineering and Computer Science (EECS), Computational Science and Technology.ORCID iD: 0000-0001-7944-4226
KTH, School of Electrical Engineering and Computer Science (EECS), Computational Science and Technology. Department of Mathematics - Stockholm University, Sweden.ORCID iD: 0000-0002-2358-7815
KTH, School of Electrical Engineering and Computer Science (EECS), Centres, Digital futures. KTH, School of Electrical Engineering and Computer Science (EECS), Computational Science and Technology.ORCID iD: 0000-0001-6553-823X
2023 (English)In: ESANN 2023 Proceedings - 31st European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Universite Catholique de Louvain , 2023, p. 611-616Conference paper, Published paper (Refereed)
Abstract [en]

We introduce a novel spiking neural network model for learning distributed internal representations from data in an unsupervised procedure. We achieved this by transforming the non-spiking feedforward Bayesian Confidence Propagation Neural Network (BCPNN) model, employing an online correlation-based Hebbian-Bayesian learning and rewiring mechanism, shown previously to perform representation learning, into a spiking neural network with Poisson statistics and low firing rate comparable to in vivo cortical pyramidal neurons. We evaluated the representations learned by our spiking model using a linear classifier and show performance close to the non-spiking BCPNN, and competitive with other Hebbian-based spiking networks when trained on MNIST and F-MNIST machine learning benchmarks.

Place, publisher, year, edition, pages
Universite Catholique de Louvain , 2023. p. 611-616
National Category
Bioinformatics (Computational Biology) Neurosciences Computer Sciences Computer graphics and computer vision
Identifiers
URN: urn:nbn:se:kth:diva-380151DOI: 10.14428/esann/2023.ES2023-169Scopus ID: 2-s2.0-105034968640OAI: oai:DiVA.org:kth-380151DiVA, id: diva2:2058020
Conference
31st European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2023, Bruges, Belgium, Oct 4 2023 - Oct 6 2023
Note

QC 20260506

Available from: 2026-05-06 Created: 2026-05-06 Last updated: 2026-09-07Bibliographically approved
In thesis
1. Brain-like Representation Learning and Associative Memory
Open this publication in new window or tab >>Brain-like Representation Learning and Associative Memory
2026 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

The brain enables organisms to perceive the world, learn from experience, generate complex behavior, and give rise to cognition. Understanding the information processing principles underlying brain computation remains a key challenge in computational neuroscience and cognitive science. Elucidating these principles has the potential to provide a foundation for developing intelligent machines and energy-efficient, scalable, and robust artificial intelligence paradigms.

This thesis investigates neural network models incorporating key brain-like design principles, focusing on unsupervised representation learning and the formation of robust associative memory. While modern deep learning approaches achieve strong performance in representation learning tasks, they rely on backpropagation-based optimization, which lacks biological plausibility and depends on globally coordinated computations. There remains a need for neural network models that can demonstrate complex functionalities while remaining grounded in biological principles.

To address this gap, this thesis develops a class of brain-like models based on the Bayesian Confidence Propagation Neural Network (BCPNN) framework. The proposed models incorporate key brain-like design principles, including localized Hebbian synaptic plasticity, structural plasticity, activity and connection sparsity, and a modular architecture derived from neocortical columnar organization. Furthermore, the models integrate feedforward, recurrent, and feedback connectivity to support both representation learning and associative memory.

This work comprises three main lines of investigation. First, the unsupervised representation learning capabilities of a feedforward model are examined, demonstrating that structured internal representations can be learned directly from unlabeled data using localized synaptic and structural plasticity. Second, the formation of associative memory is studied by integrating recurrent connectivity in the representation learning model, demonstrating robust recall from partial, noisy, or corrupted inputs when tested on pattern completion, prototype extraction, and noise robustness tasks. Third, the framework is extended to spiking neural networks, where neurons communicate via stochastic spike events, and the results show that, through synaptic short-term filtering and appropriate temporal scaling, the spiking models approximate the behavior of their rate-based counterparts while preserving functionality and performance. 

Overall, the results establish that brain-like design principles can support scalable representation learning and robust associative memory, demonstrating a bridge between neuroscience and artificial intelligence within the emerging field of NeuroAI. This work further offers a pathway toward energy-efficient, brain-like neuromorphic systems capable of operating in real-time and in dynamic environments.

Abstract [sv]

Hjärnan gör det möjligt för organismer att uppfatta världen, lära av erfarenhet, generera komplext beteende och ge upphov till kognition. Att förstå de informationsbearbetningsprinciper som ligger till grund för hjärnans beräkningar är fortfarande en central utmaning inom beräkningskognitiv neurovetenskap. Att klarlägga dessa principer har potential att utgöra en grund för utvecklingen av intelligenta maskiner samt energieffektiva, skalbara och robusta paradigm för artificiell intelligens. 

Denna avhandling undersöker neurala nätverksmodeller som införlivar centrala hjärninspirerade designprinciper, med fokus på representationsinlärning från oövervakad data och bildandet av robust associativt minne. Även om moderna djupa neurala nätverk uppvisar stark prestanda i representationsinlärningsuppgifter, förlitar de sig på optimering baserad på backpropagation, vilket saknar biologisk plausibilitet och är beroende av globalt koordinerade beräkningar. Det finns därför ett behov av neurala nätverksmodeller som kan uppvisa komplex funktionalitet samtidigt som de förblir förankrade i biologiska principer. 

För att adressera detta utvecklar denna avhandling en klass av hjärninspirerade modeller baserade på ramverket Bayesian Confidence Propagation Neural Network (BCPNN). De föreslagna modellerna införlivar centrala hjärninspirerade designprinciper, inklusive lokal Hebbsk synaptisk plasticitet, strukturell plasticitet, gleshet samt en modulär arkitektur härledd från neokortex kolumnära organisation. Vidare integrerar modellerna framåtriktade, rekurrenta och återkopplande kopplingar för att stödja både representationsinlärning och associativt minne. 

Arbetet omfattar tre huvudsakliga undersökningslinjer. För det första studeras den oövervakade representationsinlärningsförmågan hos en framåtriktad modell, vilket visar att strukturerade interna representationer kan läras direkt från oetiketterad data med hjälp av lokal synaptisk och strukturell plasticitet. För det andra undersöks bildandet av associativt minne genom att integrera rekurrenta kopplingar i representationsinlärningsmodellen, vilket demonstrerar robust återkallelse från partiella, brusiga eller störda indata vid tester av mönsterkomplettering, prototyputvinning och brusrobusthet. För det tredje utvidgas ramverket till spikande neurala nätverk, där neuroner kommunicerar via stokastiska spike-händelser, och det visas att de spikande modellerna, genom synaptisk korttidsfiltrering och lämplig tidsmässig skalning, approximerar beteendet hos sina rate-baserade motsvarigheter samtidigt som funktionalitet och prestanda bevaras. 

Sammantaget visar resultaten att hjärninspirerade designprinciper kan stödja skalbar representationsinlärning och robust associativt minne, och därmed etablera en brygga mellan neurovetenskap och artificiell intelligens inom det framväxande området NeuroAI. Detta arbete erbjuder vidare en väg mot energieffektiva, hjärninspirerade neuromorfa system som kan operera i realtid och i dynamiska miljöer.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2026. p. 121
Series
TRITA-EECS-AVL ; 2026:76
Keywords
brain-like computing, NeuroAI, neuromorphic, neural networks, representation learning, associative memory, pattern recognition, synaptic plasticity, structural plasticity, spiking neurons, Bayesian Confidence Propagation Neural Network (BCPNN), Hebbian learning, cortical columns, hjärninspirerad beräkning, NeuroAI, neuromorf, neurala nätverk, representationsinlärning, associativt minne, mönsterigenkänning, synaptisk plasticitet, strukturell plasticitet, spikande neuroner, Bayesian Confidence Propagation Neural Network (BCPNN), Hebbiansk inlärning, kortikala kolumner
National Category
Artificial Intelligence Bioinformatics (Computational Biology)
Research subject
Computer Science
Identifiers
urn:nbn:se:kth:diva-387982 (URN)978-91-8106-695-1 (ISBN)
Public defence
2026-10-01, https://kth-se.zoom.us/j/64100876364, F3 (Flodis), Lindstedtsvägen 26 & 28, Stockholm, 14:00 (English)
Opponent
Supervisors
Funder
Swedish e‐Science Research Center
Note

QC 20260907

Available from: 2026-09-07 Created: 2026-09-06 Last updated: 2026-09-07Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Ravichandran, Naresh BalajiLansner, AndersHerman, Pawel

Search in DiVA

By author/editor
Ravichandran, Naresh BalajiLansner, AndersHerman, Pawel
By organisation
Computational Science and TechnologyDigital futures
Bioinformatics (Computational Biology)NeurosciencesComputer SciencesComputer graphics and computer vision

Search outside of DiVA

GoogleGoogle Scholar

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 21 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf