Towards orientation invariant sensorimotor object recognition based on hierarchical temporal memory with cortical grid cells.

Simple item page
dc.contributor.advisorVan den Heever, D.en_ZA
dc.contributor.authorRoux, K.en_ZA
dc.contributor.otherStellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.en_ZA
dc.date.accessioned2021-06-23T15:10:26Z
dc.date.accessioned2021-12-22T14:12:05Z
dc.date.available2021-06-23T15:10:26Z
dc.date.available2021-12-22T14:12:05Z
dc.date.issued2021-12
dc.descriptionThesis (MEng)--Stellenbosch University, 2021.en_ZA
dc.description.abstractENGLISH ABSTRACT: Hierarchical Temporal Memory (HTM) is a framework that implements bio- logically plausible artificial intelligence by capturing key computational and architectural principles of the neocortex. This study proposes an extension to the HTM framework to support sensor orientations relative to learned allocentric object representations. The proposed mechanism enables object representations to be learned through sensorimotor sequences, and inference of these learned object representations from novel sensorimotor sequences produced by rotated objects through path integration. The model proposes that orientational selective cells are present in each column in the neocortex, and provides a biologically plausible implementation that echoes experimental measurements and fits in with theoretical predictions of previous studies.en_ZA
dc.description.abstractAFRIKAANSE OPSOMMING: Hiërargiese Tydelike Geheue (HTG) is n raamwerk wat biologies aanneem- like kunsmatige intelligensie implementeer deur fundamentele berekenings- en argitektoniese beginsels van die neokorteks vas te lê. Hierdie studie bied ’n uitbreiding van die HTG raamwerk om sensor oriëntasies ten opsigte van aangeleerde allosentriese verteenwoordigings te ondersteun. Die voorgestelde meganisme maak dit moontlik om verteenwoordigings te leer deur middel van sensormotoriese volgordes, en afleiding van hierdie verteenwoordigings van geleerde objekte uit nuwe sensormotoriese volgordes wat geproduseer word deur geroteerde objekte deur middel van padintegrasie. Die model stel voor dat oriëntasie-selektiewe selle in elke kolom in die neokorteks voorkom, en bied ’n biologies aanneemlike implementering wat eksperimentele metings weerspieël en inpas by teoretiese voorspellings van vorige studies.af_ZA
dc.description.versionMastersen_ZA
dc.format.extent115 pagesen_ZA
dc.identifier.urihttp://hdl.handle.net/10019.1/123610
dc.language.isoen_ZAen_ZA
dc.publisherStellenbosch : Stellenbosch Universityen_ZA
dc.rights.holderStellenbosch Universityen_ZA
dc.subjectSensorimotor -- Learningen_ZA
dc.subjectHierarchical Temporal Memoryen_ZA
dc.subjectOrientation invarianceen_ZA
dc.subjectGrid cellsen_ZA
dc.subjectUCTDen_ZA
dc.subjectNeocortexen_ZA
dc.subjectArtificial intelligence -- Physiologyen_ZA
dc.titleTowards orientation invariant sensorimotor object recognition based on hierarchical temporal memory with cortical grid cells.en_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
roux_orientation_2021.pdf
Size:
10.09 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Plain Text
Description:
Download
Collections
Masters Degrees (Mechanical and Mechatronic Engineering)