Use this URL to cite or link to this record in EThOS:
Title: On the cunning of the human retina : highlighting the mesoscopic talents of its prime computational layer
Author: Dubuc, Timothee
ISNI:       0000 0004 6421 9490
Awarding Body: University of Reading
Current Institution: University of Reading
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Restricted access.
Access from Institution:
The human retina, traditionally described as the biological equivalent of a camera, is supporting the processes involved in early visual function. Beyond its role of integrator of the photonic signals coming from the surrounding environment, one emerging hypothesis posits that it is, in fact, responsible for some more complex computations and the extraction of salient features that emerge from the interaction of the visual system with the environment. In the work presented in this thesis, we tasked ourselves to explore the potential capabilities of this piece of wetware, with a view of confronting evidence that the retina should be treated as a piece 0/ brain. We first tried to assert the presence of qualitative computations, useful to visual function, as early as the first layers of the hierarchical retinal network. We thus carried modelling efforts spanning from the microscopic to the mesoscopic levels of these early systems. This work resulted, on one hand, in the creation of light, qualitative, computational representations of the two cells involved, rod photoreceptors and horizontal cells, which we coupled with a behavioural parameter space analysis technique highlighting the handles that will be fit to drive the ensuing computations. On the other hand, we interconnected the retinal cells modelled in a position-realistic, bio-realistic network, with a view to studying and reproducing the photoreceptive retinal substrate in naturalistic conditions. We used self-organisation techniques to grow a virtual piece of retina, which permitted the study of both the behaviour exhibited by variations of the model and the dynamics supporting the qualitative computation we elected to study. Through the lens of three visual qualitative features, that we labelled "edge detection", "contextual hallucination", and "blur generation", we aim to show that even such a simple neural network is capable, assuming the right parametric conditions, of exhibiting useful sequential, transient computations, coupled with the manipulation of potentially informative multiplexed signal. Despite the biological shortcuts operated for the sake of computational tractability, we are confident that the experiments carried out constitute a step towards revealing what the human retina is actually about.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available