Analog implementation of ANN with inherent quadratic nonlinearity of the synapses

In real-life applications of multilayer neural networks, the scale of integration, processing speed, and manufacturability are of key importance. A simple analog-signal synapse model is implemented on a standard 0.35 /spl mu/m CMOS process requiring no floating-gate capability. A neural-matrix of 2176 analog current-mode synapses arranged in eight layers of 16 neurons with 16 inputs each is constructed for the purpose of a fingerprint feature extraction application. Synapse weights are stored on the analog storage capacitors, and synapse nonlinearity with respect to weight is investigated. The capability of the synapse to operate in feedforward and learning modes is studied and demonstrated. The effect of the synapse's inherent quadratic nonlinearity on learning convergence and on the optimization of vector direction is analyzed. Transistor-level analog simulations verify the hardware circuit. System-level MatLab simulations verify the synapse mathematical model. The conclusion reached is that the proposed implementation is very suitable for large-scale artificial neural networks - especially if on-chip integration with other products on a standard CMOS process is required.     

Rod milling and thermal annealing of graphite: Passing the equilibrium barrier

The change in graphitic carbon structure induced by mechanical milling has been monitored by Raman spectroscopy, transmission electron microscopy (TEM) and X-ray diffraction. It is well known that progressive rod milling of graphite results in an increase in structural disorder. Here, it has been found that a milling time of around 80 h is crucial in producing maximum nanocrystallite formation and this affects the nature of the products formed before or after annealing. At about 80 h equilibrium forms and no further production of nanocrystallites is possible although if additional energy is added amorphous carbon begins to form. Annealing produces different nanographitic carbons depending on the milling conditions because the material may be milled to an equilibrium concentration of nanocrystallites or less, or with additional energy transformed further past equilibrium to new product. Linear morphological structures and trace amounts of carbon nanotubes were found on milling for 80 h and annealing, but concentric layers of carbons were observed in samples milled as long as 240 h.     

The fibrinogen-like globe of tenascin-C mediates its interactions with neurocan and phosphacan/protein-tyrosine phosphatase-zeta/beta

Two nervous tissue-specific chondroitin sulfate proteoglycans, neurocan and phosphacan (the extracellular domain of protein-tyrosine phosphatase-zeta/beta), are high-affinity ligands of tenascin-C. Using portions of tenascin-C expressed as recombinant proteins in human fibrosarcoma cells, we have demonstrated both by direct radioligand binding assays and inhibition studies that phosphacan binding is retained in all deletion variants except those lacking the fibrinogen-like globe and that phosphacan binds to this single domain with nearly the same affinity (Kd approximately 12 nM) as to native or recombinant tenascin-C. However, maximum binding of neurocan requires both the fibrinogen globe and some of the adjacent fibronectin type III repeats. Binding of phosphacan and neurocan to intact tenascin-C, and of phosphacan to the fibrinogen globe, is significantly increased in the presence of calcium. Chondroitinase treatment of the proteoglycans did not affect their binding to either native tenascin-C or to any of the recombinant proteins, demonstrating that these interactions are mediated by the proteoglycan core proteins rather than through the glycosaminoglycan chains. These results are also consistent with rotary shadowing electron micrographs that show phosphacan as a rod terminated at one end by a globular domain that is frequently seen apposed to the fibrinogen globe in mixtures of phosphacan and tenascin-C. C6 glioma cells adhere to and spread on deletion variants of tenascin-C containing only the epidermal growth factor-like domains or the fibronectin type III repeats and the fibrinogen globe. In both cases cell adhesion was inhibited by similar concentrations of phosphacan, demonstrating that the fibrinogen globe is not necessary for this effect, which is apparently mediated by a direct action of phosphacan on the cells rather than by its interaction with the proteoglycan binding site on tenascin-C.     

Differences in serum cytokine levels in acute and chronic autoimmune thrombocytopenic purpura: relationship to platelet phenotype and antiplatelet T-cell reactivity

Patients with both acute and chronic autoimmune thrombocytopenic purpura (AITP) have in vitro lymphocyte defects in the form of platelet-stimulated proliferation and cytokine secretion. A blinded study was performed to determine if these defects are related to serum cytokine levels and/or platelet antigen expression. Compared with controls, 53% of children with chronic AITP, but only 9% of those with acute AITP, had increased serum interleukin-2 (IL-2), interferon-gamma, and/or IL-10; however, none of the patients had detectible serum levels of IL-4 or IL-6, cytokine patterns suggesting and early CD4+ Th0 and Th1 cell activation. In children with chronic AITP, the levels of serum IL-2 correlated with in vitro platelet-stimulated IL-2 production. Few (17%) patients with AITP showed platelet activation, as measured by CD62 expression, or abnormal expression levels of platelet membrane glycoprotein (GP) IIbIIIa, but abnormal GPIb levels were observed in one-third of children with AITP. In contrast to normal controls and patients with nonimmune thrombocytopenia, a significant number of children with acute (80%), chronic (71%), or chronic-complex (55%) AITP and GPIb+ peripheral blood cells expressing HLA-DR. HLA-DR was variably coexpressed on distinct smaller and larger-sized GPIb+ cell populations with CD41, CD45, CD14, CD80, and/or glycophorin molecules. GPIb+ cells isolated from spleens of patients with chronic AITP had high expression (49% +/- 30%) of HLA-DR and splenic T cells had a high level of in vitro platelet-stimulated IL-2 secretion compared with controls. Platelet HLA-DR expression correlated inversely with platelet count, but not with therapy, serum cytokines, or in vitro lymphocyte antiplatelet reactivity. The results indicate that platelet HLA-DR expression is a common occurrence in patients with immune thrombocytopenia, whereas a large subpopulation of children with chronic AITP can be identified by increased serum cytokine levels and in vitro platelet-stimulated IL-2 secretion by lymphocytes, suggesting that differences exist in the immune pathogenesis of acute and chronic AITP, particularly at the level of platelet reactive T cells.     

Dispersion of silicate nano-plates within poly(acrylic acid) and their interfacial interactions

Abstract

Dispersion of the silicate nano-plates of clays within polymers for nanocomposite formation generally involved interactions of the polymers with the plate surfaces. In this paper, we provide preliminary evidence showing dependency of the dispersion on the interactions of the plate lattices with polymers. Experiments have been carried out on the dispersion of the natural clay, montmorillonite (≈0.5 g) within the aqueous solutions of poly(acrylic acid) at varying temperatures and the products analysed by X-ray diffraction and related techniques. In the product from reaction at 60 ?C, the silicate plates are dispersed with fully extended chains of poly(acrylic acid) intercalated within the interlayer spaces between unexfoliated plates. Photoemission spectroscopy showed that during the process sodium ions are removed from the silicate surface. In the product from reaction at 85 ?C or above, the silicate plates are partly exfoliated. During the process, Fe²⁺ ions within the clay lattice are oxidised by the acidic poly(acrylic acid) solution, which suggests strong reaction of poly(acrylic acid) with the silicate plates facilitating exfoliation. Interestingly, the Al and Mg concentrations in the lattice remain virtually unchanged during the reactions.