Thesis (Ph. D.)--University of Chicago, Committee on Neurobiology, August 1999. / Includes bibliographical references. Also available on the Internet.

Thesis (Ph. D.)--University of Washington, 2002. / Vita. Includes bibliographical references (leaves 112-132).

Thesis (M.S.)--State University of New York at Binghamton, Department of Systems Science and Industrial Engineering, Thomas J. Watson School of Engineering and Applied Science, 2007. / Includes bibliographical references.

Thesis (M.A.) in Mathematics--University of Maine, 2004. / Includes vita. Includes bibliographical references (leaves 40-42).

No description available.

Precise point-to-point connectivity is the basis of ordered maps of the visual field in the brain. One point in the visual field is represented at one locus in the dLGN and one locus in primary visual cortex. A fundamental problem in the development of most sensory systems is the creation of the topographic projections which underlie these maps. Mechanisms ranging from ordered ingrowth of fibres, through chemical guidance of axons to sculpting of the map from an early exuberant input have been proposed. However, we know little about how ordered maps are created beyond the first relay. What we do know is that a topological mismatch requires the exchange of neighbours in the geniculo-cortical projection and that manipulating the input to the primary relay can affect the geniculo-cortical topography. Taking advantage of the immaturity of the newborn hamster’s visual system, I studied the generation of an ordered map in primary visual cortex during the time of target innervation in normal and manipulated animals. I also investigated the patterning of neuronal activity prior to natural eye-opening. Paired injections of retrograde fluorescent tracers into visual cortex reveal that geniculate fibres are highly disordered at the time of invasion of the cortical plate. Topography in the geniculo-cortical projection emerges out of an unordered projection to area 17 in the first postnatal week. Furthermore, I show that manipulating the peripheral input can alter the topographic map which arises out of the early scatter. Removal of one eye at birth appears to slow the process of geniculo-cortical map formation ipsilateral to the remaining eye and at the end of the second postnatal week, a double projection between thalamus and cortex has formed. If retinal activity is blocked during this time, this double projection does not emerge. The results implicate retinal activity as the signal that induces the development of a different topographic order in the geniculo-cortical projection. It is generally believed that visual experience can influence development only after eye-opening. However, the final part of my thesis shows that neurons in the developing visual cortex of the ferret can not only be visually driven at least 10 days before natural eye-opening, but are also selective for differently oriented gratings presented <i>through the closed eye-lid</i>. Thus, visually-driven neuronal activity could influence development much earlier than previously assumed in many developmental studies.

Reinforcement learning algorithms comprise a class of learning algorithms for neural networks. Reinforcement learning is distinguished from other classes by the type of problems that it is intended to solve. It is used for learning input-output mappings where the desired outputs are not known and only a scalar reinforcement value is available. Primary Reinforcement Learning (PRL) is a core component of the most actively researched form of reinforcement learning. The issues surrounding the convergence characteristics of PRL are considered in this thesis. There have been no convergence proofs for any kind of networks learning under PRL. A convergence theorem is proved in this thesis, showing that under some conditions, a particular reinforcement learning algorithm, the A[formula omitted] algorithm, will train a single-layer network correctly. The theorem is demonstrated with a series of simulations. A new PRL algorithm is proposed to deal with the training of multiple layer, binary output networks with continuous inputs. This is a more difficult learning problem than with binary inputs. The new algorithm is shown to be able to successfully train a network with multiple outputs when the environment conforms to the conditions of the convergence theorem for a single-layer network. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

The effects of the inhibitory neurotransmitters gammaamino butyric acid (GABA) and glycine were characterized on spontaneous activity recorded from mouse spinal cord cultures. The GABA concentration which completely inhibited burst activity was chosen as a quantifiable measure of culture drug response and was used to 1) assess interculture and intraculture variability, 2) determine the influence of culture age and initial activity on GABA responses, and 3) compare the GABA responses between networks obtained from whole spinal cord and ventral half spinal cord. Results showed that 1) no significant variability existed either within or among cultures, 2) the initial culture activity directly affected GABA responses, 3) the culture age had no effect on GABA responses, and 4) there was no significant difference in GABA responses between the two spinal cord tissues.

Understanding the mapping between stimulus, behavior, and neural responses is vital for understanding sensory, motor, and general neural processing. We can examine this relationship through the complementary methods of encoding (predicting neural responses given the stimulus) and decoding (reconstructing the stimulus given the neural responses). The work presented in this thesis proposes, evaluates, and analyzes several nonlinear approaches for encoding and decoding that leverage recent advances in machine learning to achieve better accuracy. We first present and analyze a recurrent neural network encoding model to predict retinal ganglion cell responses to natural scenes, followed by a decoding approach that uses neural networks for approximate Bayesian decoding of natural images from these retinal cells. Finally, we present a probabilistic framework to distill behavioral videos into useful low-dimensional variables and to decode this behavior from neural activity.

Neural circuits within the hippocampus, a mammalian brain structure critical for both the encoding and consolidation of episodic memories, are composed of intimately connected excitatory pyramidal cells and inhibitory interneurons. While decades of research have focused on how the in vivo physiological properties of pyramidal cells may support these cognitive processes, and the anatomical and physiological properties of interneurons have been extensively studied in vitro, relatively little is known about how the in vivo activity patterns of interneurons support memory encoding and consolidation. Here, I have utilized Acousto-Optic Deflection (AOD)-based two-photon calcium imaging and post-hoc immunohistochemistry to perform large-scale recordings of molecularly-defined interneuron subtypes, within both CA1 and CA3, during various behavioral tasks and states. I conclude that the subtype-specific dynamics of inhibitory circuits within the hippocampus are critical in supporting its role in memory encoding and consolidation.