Visualization of learning in multilayer perceptron networks using principal component analysis

This paper is concerned with the use of scientific visualization methods for the analysis of feedforward neural networks (NNs). Inevitably, the kinds of data associated with the design and implementation of neural networks are of very high dimensionality, presenting a major challenge for visualization. A method is described using the well-known statistical technique of principal component analysis (PCA). This is found to be an effective and useful method of visualizing the learning trajectories of many learning algorithms such as back-propagation and can also be used to provide insight into the learning process and the nature of the error surface.

Visuospatial processing and the function of prefrontal-parietal networks in autism spectrum disorders: a functional MRI study

Objective: Individuals with autism spectrum disorders typically have normal visuospatial abilities but impaired executive functioning, particularly in abilities related to working memory and attention. The aim of this study was to elucidate the functioning of frontoparietal networks underlying spatial working memory processes during mental rotation in persons with autism spectrum disorders. Method: Seven adolescent males with normal IQ with an autism spectrum disorder and nine age- and IQ-matched male comparison subjects underwent functional magnetic resonance imaging scans while performing a mental rotation task. Results: The autism spectrum disorders group showed less activation in lateral and medial premotor cortex, dorsolateral prefrontal cortex, anterior cingulate gyrus, and caudate nucleus. Conclusions: The finding of less activation in prefrontal regions but not in parietal regions supports a model of dysfunction of frontostriatal networks in autism spectrum disorders.

The generation of H-1-NMR-detectable mobile lipid in stimulated lymphocytes: Relationship ts cellular activation, the cell cycle, and phosphatidylcholine-specific phospholipase C

Mobile Lipids detected using H-1-NMR in stimulated lymphocytes were correlated with cell cycle phase, expression of the interleukin-2 receptor alpha and proliferation to assess the activation status of the lymphocytes. Mobile lipid levels, IL-2R alpha expression and proliferation increased after treatment with PMA and ionomycin. PMA or ionomycin stimulation alone induced increased IL-2R alpha expressiom but not proliferation, PMA- but not ionomycin-stimulation generated mobile lipid, Treatment with anti-CD3 antibody did not increase IL-2R alpha expression or proliferation but did generate increased amounts of mobile lipid, The cell cycle status of thymocytes treated with anti-CD3, PMA or ionomycin alone indicated an. accumulation of the cells in the G(1) phase of the cell cycle, The generation of mobile lipid was abrogated in anti-CD3 antibody-stimulated thymic lymphocytes but not in splenic lymphocytes, using a phosphatidylcholine-specific phospholipase C (PC-PLC) inhibitor which blocked cells in the G(1)/S phase of the cell cycle, This suggests that the H-1-NMR-detectable mobile Lipid may be generated in anti-CD3 antibody-stimulated thymic lymphocytes by the action of PC-PLC activity via the catabolism of PC, in the absence of classical signs of activation. (C) 1997 Academic Press.

Retinal neurons: cell types and coupled networks

Retinal neurons with distinct dendritic morphologies are likely to comprise different cell types, subject to three important caveats. First, it is necessary to avoid creating “artificial” cell types based on arbitrary criteria—for example, the presence of two or three primary dendrites. Second, it is essential to take into account changes in morphology with retinal eccentricity and cell density. Third, the retina contains imperfections like any natural system and a significant number of retinal neurons display aberrant morphologies or make aberrant connections that are not typical of the population as a whole. Many types of retinal ganglion cells show diverse patterns of tracer coupling, with the simplest pattern represented by the homologous coupling shown by On-Off direction-selective (DS) ganglion cells in the rabbit retina. Neighboring DS ganglion cells with a common preferred direction have regularly spaced somata and territorial dendritic fields, whereas DS ganglion cells with different preferred directions may have closely spaced somata and overlapping dendritic fields.