In this thesis, a method is proposed with which good modular artificial neural network structures can be found automatically using a computer program. A number of biological metaphors are incorporated in the method. It will be argued that modular artificial neural networks have a better performance than their non-modular counterparts. The human brain can also be seen as a modular neural network, and the proposed search method is based on the natural process that resulted in the brain: Genetic algorithms are used to imitate evolution, and L-systems are used to model the kind of recipes nature uses in biological growth. A small number of experiments have been done to investigate the possibilities of the method. Preliminary results show that the method does find modular networks, and that those networks outperform `standard' solutions. The method looks very promising, although the experiments done were too limited to draw any general conclusions. One drawback is the large amount of compu...

Many researchers interested in connectionist models accept that such models are "neurally inspired " but do not worry too much about whether their models are biologically realistic. While such a position may be perfectly justifiable, the present paper attempts to illustrate how biological information can be used to constrain connectionist models. Two particular areas are discussed. The first section deals with visual information processing in the primate and human visual system. It is argued that speed with which visual information is processed imposes major constraints on the architecture and operation of the visual system. In particular, it seems that a great deal of processing must depend on a single bottum-up pass. The second section deals with biological aspects of learning algorithms. It is argued that although there is good evidence for certain coactivation related synaptic modification schemes, other learning mechanisms, including back-propagation, are not currently supported by experimental data.

... This article models how these interactions help visual cortex to realize: (1) the binding process whereby cortex groups distributed data into coherent object representations; (2) the attentional process whereby cortex selectively processes important events; and (3) the developmental and learning processes whereby cortex shapes its circuits to match environmental constraints. New computational ideas about feedback systems suggest how neocortex develops and learns in a stable way, and why top-down attention requires converging bottom-up inputs to fully activate cortical cells, whereas perceptual groupings do not.

Using noninvasive functional magnetic resonance imaging (fMRI) techniques, we analyzed the responses in human area MT with regard to visual motion, color, and luminance contrast sensitivity, and retinotopy. As in previous PET studies, we found that area MT responded selectively to moving (compared to stationary) stimuli. The location of human MT in the present fMRl results is consistent with that of MT in earlier PET and anatomical studies. In addition we found that area MT has a much higher contrast sensitivity than that in several other areas, includ-ing primary visual cortex (Vl). Functional MRI half-ampli-tudes in Vl and MT occurred at approximately 15 % and 1% luminance contrast, respectively. High sensitivity to con-trast and motion in MT have been closely associated with magnocellular stream specialization in nonhuman pri-mates. Human psychophysics indicates that visual motion ap-pears to diminish when moving color-varying stimuli are equated in luminance. Electrophysiological results from macaque MT suggest that the human percept could be due to decreases in firing of area MT cells at equiluminance. We show here that fMRl activity in human MT does in fact decrease at and near individually measured equilumi-nance. Tests with visuotopically restricted stimuli in each hem-ifield produced spatial variations in fMRl activity consistent with retinotopy in human homologs of macaque areas Vl, V2, V3, and VP. Such activity in area MT appeared much less retinotopic, as in macaque. However, it was possible to measure the interhemispheric spread of fMRl activity in human MT (half amplitude activation across the vertical meridian =-15’).

We have used positron emission tomography (PET), which measures regional cerebral blood flow (rCBF), to demonstrate directly the specialization of function in the normal human visual cortex. A novel technique, statistical parametric mapping, was used to detect foci of significant change in cerebral blood flow within the prestriate cortex, in order to localize those parts involved in the perception of color and visual motion. For color, we stimulated the subjects with a multicolored abstract display containing no recognizable objects (Land color Mondrian) and contrasted the resulting blood flow maps with those obtained when subjects viewed an identical display consisting of equiluminous shades of gray. The comparison identified a unique area (area V4) located in the lingual and fusiform gyri of the prestriate cortex. For motion, blood flow maps when subjects viewed moving

Abstract not found

Abstract not found

An intelligent system has to be capable of adapting to a constantly changing environment. It therefore, ought to be capable of learning from its perceptual interactions with its surroundings. This requires a certain amount of plasticity in its structure. Any attempt to model the perceptual capabilities of a living system or, for that matter, to construct a synthetic system of comparable abilities, must therefore, account for such plasticity through a variety of developmental and learning mechanisms. This paper examines some results from neuroanatomical, morphological, as well as behavioral studies of the development of visual perception; integrates them into a computational framework; and suggests several interesting experiments with computational models that can yield insights into the development of visual perception. Role of Environmental Experience in Development and Learning In order to understand the development of information processing structures in the brain, one needs knowl...

Abstract not found

The visual brain consists of several parallel, functionally specialized processing systems, each having several stages (nodes) which terminate their tasks at different times; consequently, simultaneously presented attributes are perceived at the same time if processed at the same node and at different times if processed by different nodes. Clinical evidence shows that these processing systems can act fairly autonomously. Damage restricted to one system compromises specifically the perception of the attribute that that system is specialized for; damage to a given node of a processing system that leaves earlier nodes intact results in a degraded perceptual capacity for the relevant attribute, which is directly related to the physiological capacities of the cells left intact by the damage. By contrast, a system that is spared when all others are damaged can function more or less normally. Moreover, internally created visual percepts—illusions, afterimages, imagery, and hallucinations—activate specifically the nodes specialized for the attribute perceived. Finally, anatomical evidence shows that there is no final integrator station in the brain, one which receives input from all visual areas; instead, each node has multiple outputs and no node is recipient only. Taken together, the above evidence leads us to propose that each node of a processing-perceptual system creates its own microconsciousness. We propose that, if any binding occurs to give us our integrated image of the visual world, it must be a binding between microconsciousnesses generated at different nodes. Since any two microconsciousnesses generated at any two nodes can be bound together, perceptual integration is not hierarchical, but parallel and postconscious. By contrast, the neural machinery conferring properties on those cells whose activity has a conscious correlate is hierarchical, and we refer to it as generative binding, to distinguish it from the binding that might occur between the microconsciousnesses. © 1999 Academic Press