Show simple item record

dc.contributor.authorAlbert, Marken_US
dc.date.accessioned2010-04-09T20:21:44Z
dc.date.available2015-04-09T06:27:42Z
dc.date.issued2010-04-09T20:21:44Z
dc.identifier.otherbibid: 6890922
dc.identifier.urihttps://hdl.handle.net/1813/14789
dc.description.abstractTraditionally, visual development is thought to occur in two distinct stages, an innate stage which occurs before eye-opening and a learning stage in which experience shapes development. This dissertation will show how the traditional view represents a false dichotomy limiting progress in understanding visual development. Prior to eye opening there is patterned, spontaneous neural activity. Models will illustrate that spontaneous neural activity can shape development in the same way as early, natural visual experience; the same learning method can be used in both stages of development. A parsimonious view of early development is proposed - a method of 'innate learning' that prepares the visual system for later experienced-based refinement by learning from endogenous, patterned neural activity. This dissertation presents a series of research studies which uniquely apply a normative, efficient coding approach to understand visual development prior to eye opening. The neural code in fully-developed adult primary visual cortex (V1) is presumed to be efficient. By efficiently encoding images of our natural visual environment, techniques such as sparse coding and independent components analysis (ICA) have produced linear filters which resemble experimentally measured receptive fields in V1. Simulations in this dissertation show how spontaneous patterns of activity are capable of producing these V1-like visual codes using the same learning algorithms. The spontaneous activity models created here are abstractions of known retinal, LGN, and V1 physiology; they relate mathematically to simpler theoretical models used in the physics of critical phenomena (e.g. percolation networks) but have parameters with direct physiological interpretations. Unlike previous models of V1 development, the innate learning principle applied here will be shown to constrain the statistical structure and form of modeled spontaneous activity to more closely resemble measured spontaneous activity patterns. The monocular activity model is later extended to include binocular neural activity; binocular disparity selective cells are refined by spontaneous activity correlated between eye-specific neurons in the LGN and V1 to promote early stereopsis. The implications of this normative approach to visual development will be explored in detail, showing how this 'innate learning' approach can greatly improve our understanding of sensory development.en_US
dc.language.isoen_USen_US
dc.titleNormative Visual Development: Innate Learning In The Early Visual Systemen_US
dc.typedissertation or thesisen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record

Statistics