Uncertainty is inherent in neural processing due to noise in sensation and the sensory transmission processes, the ill-posed nature of many perceptual tasks, and temporal dynamics of the natural environment, to name a few causes. A wealth of evidence from physiological and behavioral experiments show that these various forms of uncertainty have major effects on perceptual learning and inference. In order to use sensory inputs efficiently to make decisions and guide behavior, neural systems must represent and manipulate information about uncertainty in their computations. In this thesis, we first consider how spiking neural populations might encode and decode information about continuous dynamic stimulus variables including the uncertainty about them. We explore the efficacy of a complex encoder that is paired with a simple decoder which allows computationally straightforward representation and manipulation of dynamically changing uncertainty. The encoder we present takes the form of a biologically plausible recurrent spiking neural network where the output population recodes its inputs to produce spikes that are independently decodeable. We show that this network can be learned in a supervised manner, by a simple, local learning rule. We also demonstrate that the coding scheme can be applied recursively to carry out meaningful uncertainty-sensitive computations such as dynamic cue combination. Next, we explore the computational principles that underlie non-linear response characteristics such as perceptual bias and uncertainty observed in audiovisual spatial illusions that involve multisensory interactions with conflicting cues. We examine in detail, the explanatory power of one particular causal model in characterizing the impact of conflicting inputs on perception and behavior. We also attempt to understand from a computational perspective, whether and how different task instructions might modulate the interaction of information from individual (visual and auditory) senses. Our analyses reveal some new properties of the sensory likelihoods and stimulus prior which were thought to be well described by Gaussian functions. Our results conclude that task-specific expectations can influence perception in ways that relate to a choice of inference strategy.

Uncertainty is inherent in neural processing due to noise in sensation and the sensory transmission processes, the ill-posed nature of many perceptual tasks, and temporal dynamics of the natural environment, to name a few causes. A wealth of evidence from physiological and behavioral experiments show that these various forms of uncertainty have major effects on perceptual learning and inference. In order to use sensory inputs efficiently to make decisions and guide behavior, neural systems must represent and manipulate information about uncertainty in their computations. In this thesis, we first consider how spiking neural populations might encode and decode information about continuous dynamic stimulus variables including the uncertainty about them. We explore the efficacy of a complex encoder that is paired with a simple decoder which allows computationally straightforward representation and manipulation of dynamically changing uncertainty. The encoder we present takes the form of a biologically plausible recurrent spiking neural network where the output population recodes its inputs to produce spikes that are independently decodeable. We show that this network can be learned in a supervised manner, by a simple, local learning rule. We also demonstrate that the coding scheme can be applied recursively to carry out meaningful uncertainty-sensitive computations such as dynamic cue combination. Next, we explore the computational principles that underlie non-linear response characteristics such as perceptual bias and uncertainty observed in audiovisual spatial illusions that involve multisensory interactions with conflicting cues. We examine in detail, the explanatory power of one particular causal model in characterizing the impact of conflicting inputs on perception and behavior. We also attempt to understand from a computational perspective, whether and how different task instructions might modulate the interaction of information from individual (visual and auditory) senses. Our analyses reveal some new properties of the sensory likelihoods and stimulus prior which were thought to be well described by Gaussian functions. Our results conclude that task-specific expectations can influence perception in ways that relate to a choice of inference strategy.

A consensus on how to characterise the anterograde and retrograde memory processes that are lost or spared after hippocampal damage has not been reached. In this thesis, I critically re-examine the empirical literature and the assumptions behind current theories. I formulate a coherent view of what makes a task hippocampally dependent at acquisition and how this relates to its long-term fate. Findings from a neural net simulation indicate the plausibility of my proposals. My proposals both extend and constrain current views on the role of the hippocampus in the rapid acquisition of information and in learning complex associations. In general, tasks are most likely to require the hippocampus for acquisition if they involve rapid, associative learning about unfamiliar, complex, low salience stimuli. However, none of these factors alone is sufficient to obligatorily implicate the hippocampus in acquisition. With the exception of associations with supra-modal information that are always dependent on the hippocampus, it is the combination of factors that is important. Detailed, complex information that is obligatorily hippocampally-dependent at acquisition remains so for its lifetime. However, all memories are semanticised as they age through the loss of detailed context-specific information and because generic cortically-represented information starts to dominate recall. Initially hippocampally dependent memories may appear to become independent of the hippocampus over time, but recall changes qualitatively. Multi-stage, lifelong post-acquisition memory processes produce semanticised re-representations of memories of differing specificity and complexity, that can serve different purposes. The model simulates hippocampal and cortical interactions in the acquisition and maintenance of episodic and semantic events, and behaves in accordance with my proposals. In particular, conceptualising episodic and semantic memory as representing points on a continuum of memory types appears viable. Support is also found for proposals on the relative importance of the hippocampus and cortex in the rapid acquisition of information and the acquisition of complex multi-model information; and the effect of existing knowledge on new learning. Furthermore, episodic and semantic events become differentially dependent on cortical and hippocampal components. Finally, as a memory ages, it is automatically semanticised and becomes cortically dependent.

Stereopsis is a par excellence demonstration of the computational power that neural systems can encapsulate. How is the brain capable of swiftly transforming a stream of binocular two-dimensional signals into a cohesive three-dimensional percept? Many brain regions have been implicated in stereoscopic processing, but their roles remain poorly understood. This dissertation focuses on the contributions of primary and dorsomedial visual cortex. Using convolutional neural networks, we found that disparity encoding in primary visual cortex can be explained by shallow, feed-forward networks optimized to extract absolute depth from naturalistic images. These networks develop physiologically plausible receptive fields, and predict neural responses to highly unnatural stimuli commonly used in the laboratory. They do not necessarily relate to our experience of depth, but seem to act as a bottleneck for depth perception. Conversely, neural activity in downstream specialized areas is likely to be a more faithful correlate of depth perception. Using ultra-high field functional magnetic resonance imaging in humans, we revealed systematic and reproducible cortical organization for stereoscopic depth in dorsal visual areas V3A and V3B/KO. Within these regions, depth selectivity was inversely related to depth magnitude — a key characteristic of stereoscopic perception. Finally, we report evidence for a differential contribution of cortical layers in stereoscopic depth perception.

An important problem in systems biology is the inference of biochemical pathways and regulatory networks from postgenomic data. The recent substantial increase in the availability of such data has stimulated the interest in inferring the networks and pathways from the data themselves. The main interests of this thesis are the application, evaluation and the improvement of machine learning methods applied to the reverse engineering of biochemical pathways and networks. The thesis starts with the application of an established method to newly available gene expression data related to the interferon pathway of the human immune system in order to identify active subpathways under di erent experimental conditions. The thesis continues with the comparative evaluation of various machine learning methods (Relevance networks, Graphical Gaussian Models, Bayesian networks) using observational and interventional data from cytometry experiments as well as simulated data from a gold-standard network. The thesis also extends and improves existing methods to include biological prior knowledge under the Bayesian approach in order to increase the accuracy of the predicted networks and it quanti es to what extent the reconstruction accuracy can be improved in this way.

Humans and animals can plan and execute movements much more adaptably and reliably than current computers can calculate robotic limb trajectories. Over recent decades, it has been suggested that our brains use motor primitives as blocks to build up movements. In broad terms a primitive is a segment of pre-optimised movement allowing a simplified movement planning solution. This thesis explores a generative model of handwriting based upon the concept of motor primitives. Unlike most primitive extraction studies, the primitives here are time extended blocks that are superimposed with character specific offsets to create a pen trajectory. This thesis shows how handwriting can be represented using a simple fixed function superposition model, where the variation in the handwriting arises from timing variation in the onset of the functions. Furthermore, it is shown how handwriting style variations could be due to primitive function differences between individuals, and how the timing code could provide a style invariant representation of the handwriting. The spike timing representation of the pen movements provides an extremely compact code, which could resemble internal spiking neural representations in the brain. The model proposes an novel way to infer primitives in data, and the proposed formalised probabilistic model allows informative priors to be introduced providing a more accurate inference of primitive shape and timing.

How can science possibly understand the organ through which the Universe knows itself? The scientific method can be used to study how electro-chemical signals represent information in the brain. However, modelling it by simulating its structures and functions is a computation- and communication-intensive task. Whilst supercomputers offer great computational power, brain-scale models are challenging in terms of communication overheads and power consumption. Dedicated neural hardware can be used to enhance simulation performance, but it is often optimised for specific models. While performance and flexibility are desirable simulation features, there is no perfect modelling platform, and the choice is subordinate to the specific research question being investigated. In this context SpiNNaker constitutes a novel parallel architecture, with communication and memory accesses optimised for spike-based computation, permitting simulation of large spiking neural networks in real time. To exploit SpiNNaker's performance and reconfigurability fully, a neural network model must be translated from its conceptual form into data structures for a parallel system. This thesis presents a flexible approach to distributing and mapping neural models onto SpiNNaker, within the constraints introduced by its specialised architecture. The conceptual map underlying this approach characterizes the interaction between the model and the system: during the build phase the model is placed on SpiNNaker; at runtime, placement information mediates communication with devices and instrumentation for data analysis. Integration within the computational neuroscience community is achieved by interfaces to two domain-specific languages: PyNN and Nengo. The real-time, event-driven nature of the SpiNNaker platform is explored using address-event representation sensors and robots, performing visual processing using a silicon retina, and navigation on a robotic platform based on a cortical, basal ganglia and hippocampal place cells model. The approach has been successfully exploited to run models on all iterations of SpiNNaker chips and development boards to date, and demonstrated live in workshops and conferences.

No description available.

Este trabalho realiza uma análise dos modelos pertencentes à Computação Neural e à Computação Evolutiva visando identificar semelhanças entre as áreas e sustentar mapeamentos entre as semelhanças identificadas. Neste contexto, a identificação de similaridades visando à resolução de problemas de otimização combinatorial resulta em uma comparação entre a Máquina de Boltzmann e os Algoritmos Evolutivos binários com população composta por um único indivíduo pai e um único indivíduo descendente. Como forma de auxiliar nas análises, o trabalho utiliza o Problema do Caixeiro Viajante como plataforma de ensaios, propondo mapeamentos entre as equações da Máquina de Boltzmann e os operadores evolutivos da Estratégia Evolutiva (1+1)-ES. / An analysis between the Evolutionary Computation and the Neural Computation fields was presented in order to identify similarities and mappings between the theories. In the analysis, the identification of similarities between the models designed for combinatorial optimization problems results in a comparison between the Boltzmann Machine and the Two-Membered Evolutionary Algorithms. In order to analyze the class of combinatorial optimization problems, this work used the Traveling Salesman Problem as a study subject, where the Boltzmann Machine equations were used to implement the evolutionary operators of an Evolution Strategy (1+1)-ES.

Este trabalho realiza uma análise dos modelos pertencentes à Computação Neural e à Computação Evolutiva visando identificar semelhanças entre as áreas e sustentar mapeamentos entre as semelhanças identificadas. Neste contexto, a identificação de similaridades visando à resolução de problemas de otimização combinatorial resulta em uma comparação entre a Máquina de Boltzmann e os Algoritmos Evolutivos binários com população composta por um único indivíduo pai e um único indivíduo descendente. Como forma de auxiliar nas análises, o trabalho utiliza o Problema do Caixeiro Viajante como plataforma de ensaios, propondo mapeamentos entre as equações da Máquina de Boltzmann e os operadores evolutivos da Estratégia Evolutiva (1+1)-ES. / An analysis between the Evolutionary Computation and the Neural Computation fields was presented in order to identify similarities and mappings between the theories. In the analysis, the identification of similarities between the models designed for combinatorial optimization problems results in a comparison between the Boltzmann Machine and the Two-Membered Evolutionary Algorithms. In order to analyze the class of combinatorial optimization problems, this work used the Traveling Salesman Problem as a study subject, where the Boltzmann Machine equations were used to implement the evolutionary operators of an Evolution Strategy (1+1)-ES.