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
Hippocampus as an Echo State Network
KTH, School of Electrical Engineering and Computer Science (EECS).
2018 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Hippocampus som ett Echo State Network (Swedish)
Abstract [en]

The Hippocampus is a brain region responsible for learning, memoryand spatial navigation. In that, the interactions between the CA3 andCA1 subregions have been the most studied due to the interesting dynamicsbetween the two regions. The excitatory auto-associative connectionsin the CA3 and the lack thereof in CA1 can be modelled asan Echo State Network (ESN) with the reservoir and readout approximatingCA3 and CA1 respectively. However, CA1 possesses somedegree of recurrent connections between the excitatory and the inhibitoryneurons, thereby posing an important problem from the computationaland Machine Learning perspective. The aim of this thesisis to introduce the recurrent connections in the readout and exploringits implications. By doing so, we observed that the recurrent connectionsperform a dynamic mapping of the readout output that makesthe system susceptible to noise, thereby affecting the performance.However, we also observed that by controlling certain parameters, themodel with the recurrent readout connections could perform comparablywith the basic ESN.

Abstract [sv]

Hippocampus är en hjärnregion som är involverad i inlärning, minne och navigering. För att öka vår kunskap om detta har interaktionerna mellan delregionerna CA3 och CA1 varit de mest studerade, detta på grund av den intressanta dynamik som uppstår mellan de två regionerna. De excitatoriska auto-associativa kopplingarna i CA3 samt avsaknaden av sådana i CA1 kan modelleras som ett Echo State Network (ESN) där reservoaren och readout kan mappas på CA3 respektive CA1. CA1 har emellertid en viss grad av rekurrenta kopplingar mellan de excitatoriska och de hämmande neuronerna, vilket kan ses som en utmaning att förstå utifrån ett maskinlärningsperspektiv. Syftet med detta projekt är att introducera de rekurrenta kopplingarna i readoutmodulen och utforska konsekvenserna. Vi observerade att de rekurrenta kopplingarna utför en dynamisk mappning av readout vilket gör systemet känsligt för brus och därigenom påverkar prestandan. Vi observerade emellertid också att genom att anpassa vissa parametrar kunde modellen med de rekurrenta readout-kopplingarna prestera jämförbart med ett standard-ESN.

Place, publisher, year, edition, pages
2018. , p. 56
Series
TRITA-EECS-EX ; 2019:10
National Category
Engineering and Technology
Identifiers
URN: urn:nbn:se:kth:diva-241368OAI: oai:DiVA.org:kth-241368DiVA, id: diva2:1280738
Subject / course
Computer Science
Educational program
Master of Science - Machine Learning
Presentation
2018-06-27, 04:57 (English)
Supervisors
Examiners
Available from: 2019-02-25 Created: 2019-01-21 Last updated: 2019-02-25Bibliographically approved

Open Access in DiVA

fulltext(1550 kB)51 downloads
File information
File name FULLTEXT01.pdfFile size 1550 kBChecksum SHA-512
93fc01121ee427cc236339251c10660788f836905de1cd19577e6a0eae65f2cd68d814eb6e0f4a3e0288361dafbe9c03b2acace2eec824f7e8f431e95fe28d73
Type fulltextMimetype application/pdf

Search in DiVA

By author/editor
Mutalik, Prabhanjan
By organisation
School of Electrical Engineering and Computer Science (EECS)
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 51 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 234 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