On early sensory experience in humans and machines

Please use this identifier to cite or link to this item:
https://doi.org/10.48693/526
Open Access logo originally created by the Public Library of Science (PLoS)
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorProf. Dr. Gordon Pipager
dc.creatorVogelsang, Marin-
dc.date.accessioned2024-03-07T17:13:33Z-
dc.date.available2024-03-07T17:13:33Z-
dc.date.issued2024-03-07T17:13:34Z-
dc.identifier.urihttps://doi.org/10.48693/526-
dc.identifier.urihttps://osnadocs.ub.uni-osnabrueck.de/handle/ds-2024030711011-
dc.description.abstractHuman perceptual development typically evolves in a temporally structured manner. As newborns, we start out with limited perceptual abilities and acquire greater and greater proficiencies over the subsequent months or years. In the visual modality, for instance, a newborn experiences the world with poor color sensitivity, resolution acuity, and contrast sensitivity initially. Then, over time, greater visual capabilities are being acquired. While these developmental progressions are well-established, their potential functional significance is not yet determined. In this thesis, I report computational tests of the hypothesis that initial sensory degradations characteristic of normal human development actively support the acquisition of perceptual proficiencies. In a first study, inspired by the low-frequency prenatal auditory experience of a fetus in the intrauterine environment, I studied the consequences of commencing auditory development with initially low-pass-filtered sounds. In a second study, I examined the impact of beginning visual experience with initially color-degraded inputs. In both cases, simulations with deep neural networks revealed that training with a developmentally inspired progression of inputs, evolving from poor to rich, led to superior generalization performance and the emergence of spatially or temporally extended receptive fields. As part of a third study, I conducted computational explorations of the consequences of the joint temporal progression of visual acuity and color sensitivity during early visual development. Specifically, I tested whether this joint progression may help account for the origin of an important organizational principle of the early visual system: the division into the magnocellular and the parvocellular pathway. My computational results provide support for this hypothesis. In addition, some of the computational studies are complemented by experiments with children who were born blind and treated for their blindness late in life, as part of the joint humanitarian and scientific initiative Project Prakash. In addition to providing important insights into cortical plasticity late in life, some of these results provide additional support in favor of the hypothesis that initial sensory degradations may be adaptive. Together, this work sheds new light on the functional significance of normal developmental trajectories, helps account for some perceptual deficits reported in individuals whose perceptual experience differs from normal development, and inspires new computational training procedures for deep neural networks.eng
dc.rightsAttribution-NonCommercial 3.0 Germany*
dc.rights.urihttp://creativecommons.org/licenses/by-nc/3.0/de/*
dc.subjectPerceptual developmenteng
dc.subjectDeep neural networkseng
dc.subjectVisual developmenteng
dc.subject.ddc500 - Naturwissenschaftenger
dc.titleOn early sensory experience in humans and machineseng
dc.typeDissertation oder Habilitation [doctoralThesis]-
thesis.locationOsnabrück-
thesis.institutionUniversität-
thesis.typeDissertation [thesis.doctoral]-
thesis.date2024-02-09-
dc.contributor.refereeProf. Dr. Pawan Sinhager
dc.contributor.refereeProf. Dr. Peter Königger
Appears in Collections:FB08 - E-Dissertationen

Files in This Item:
File Description SizeFormat 
thesis_vogelsang.pdfPräsentationsformat46,72 MBAdobe PDF
thesis_vogelsang.pdf
Thumbnail
View/Open


This item is licensed under a Creative Commons License Creative Commons