Neural Network Modelling with Input Uncertainty: Theory and Application

It is generally assumed when using Bayesian inference methods for neural networks that the input data contains no noise. For real-world (errors in variable) problems this is clearly an unsafe assumption. This paper presents a Bayesian neural network framework which accounts for input noise provided that a model of the noise process exists. In the limit where the noise process is small and symmetric it is shown, using the Laplace approximation, that this method adds an extra term to the usual Bayesian error bar which depends on the variance of the input noise process. Further, by treating the true (noiseless) input as a hidden variable, and sampling this jointly with the network's weights, using a Markov chain Monte Carlo method, it is demonstrated that it is possible to infer the regression over the noiseless input. This leads to the possibility of training an accurate model of a system using less accurate, or more uncertain, data. This is demonstrated on both the, synthetic, noisy sine wave problem and a real problem of inferring the forward model for a satellite radar backscatter system used to predict sea surface wind vectors.     

High efficiency,high switching speed,AlGaAs/GaAs P-HEMT DC–DC converter for integrated power amplifier modules

This paper presents a high efficiency, high switching frequency DC–DC buck converter in AlGaAs/GaAs technology, targeting integrated power amplifier modules for wireless communications. The switch mode, inductor load DC–DC converter adopts an interleaved structure with negatively coupled inductors. Analysis of the effect of negative coupling on the steady state and transient response of the converter is given. The coupling factor is selected to achieve a maximum power efficiency under a given duty cycle with a minimum penalty on the current ripple performance. The DC–DC converter is implemented in 0.5 μm GaAs p-HEMT process and occupies 2 × 2.1 mm² without the output network. An 8.7 nH filter inductor is implemented in 65 μm thick top copper metal layer, and flip chip bonded to the DC–DC converter board. The integrated inductor achieves a quality factor of 26 at 150 MHz. The proposed converter converts 4.5 V input to 3.3 V output for 1 A load current under 150 MHz switching frequency with a measured power efficiency of 84%, which is one of the highest efficiencies reported to date for similar current/voltage ratings.     

Device processing and characterisation of high temperature silicon carbide Schottky diodes

High temperature silicon carbide diodes with nickel silicide Schottky contacts were fabricated by deposition of titanium-nickel metal film on 4H-SiC epitaxial wafer followed by annealing at 550 °C in vacuum. Room temperature boron implantation have been used to form single zone junction termination extension. 4H-SiC epitaxial structures designed to have theoretical parallel-plain breakdown voltages of 1900 and 3600 V have been used for this research. The diodes revealed soft recoverable avalanche breakdown at voltages of 1450 and 3400 V, respectively, which are about 80% and 95% of theoretical values. I-V characteristics of fabricated 4H-SiC Schottky diodes have been measured at temperatures from room temperature up to 400 °C. The diodes revealed unchangeable barrier heights and ideality factors as well as positive coefficients of breakdown voltage.     

Mid-infrared ring laser

The first mid-infrared ring laser diode based on InAs and operating in the mid-infrared spectral region near 3 /spl mu/m and with a maximum operating temperature of 125 K is reported. The source is based on a symmetrical double heterostructure with large band offsets.     

Characteristic Anthocyanin Pattern from Onions and other Allium spp

Three red onion (Allium cepa) cultivars, top onion (A. cepa var. vivi-parum), A. altaicum and chive (A. schoenoprasum) contained several or all of the following anthocyanins: 3-(6′-malonyl-3′-glucosylglucoside), 3-(3′,6′-dimalonylglucoside), 3-(6′-malonylglucoside), 3-(3′-malonyl-glucoside), 3-(3′-glucosylglucoside) and 3-glucoside of cyanidin. Trace amounts of two pelargonidin derivatives and the 3,5-diglucosides of cyanidin and peonidin were for the first time reported in red onion. Carbon NMR data showed that the sugars of the 3-(6′-malonyl-3′-glu-cosylglucoside), 3-(6′-malonylglucoside) and 3-glucoside of cyanidin were pyranoses. Substitution either by sugar or acid in the sugar 3-position of an anthocyanin has never been reported outside the genus Allium.     

Actinorhodin is a chromosomally-determined antibiotic in Streptomyces coelicolar A3(2)

Streptomyces coelicolar A3(2) synthesizes a second antibiotic, in addition to the plasmid-determined methylenomycin A. It was identified, primarily on the evidence of mass spectroscopy of its diethyl ester, as actinorhodin, which has been described previously in other strains. It inhibited most Gram-positive bacteria tested, but only at a comparatively high concentration. Five independent mutations leading to lack of actinorhodin synthesis were located between cysD and strA on the chromosome.     

Preparation of a crystalline tris-(N-methylpyridinium)-derivative of bovine insulin

The reaction between 2-chloro-N-methylpyridinium iodide and bovine insulin at pH 8.5 afforded a trisubstituted derivative as main product. Following purification on DEAE-Sephadex A-25, this derivative crystallized from a citrate buffer in the form of large rhombohedra very similar to native insulin. This observation was confirmed by zero-layer X-ray precession photography. Amino acid analysis indicated substitution at the B1-terminus and on two of the four tyrosine residues. In the mouse convulsion assay the derivative showed no biological activity.     

Stimulation of glycogenolysis by beta adrenergic agonists in skeletal muscle of mice with the phosphorylase kinase deficiency mutation (I strain)

The mechanism by which beta adrenergic agonist stimulate glycogenolysis in intact skeletal muscle was investigated in mice with the phosphorylase kinase deficiency mutation (I strain). Although extracts of I strain diaphragm muscle had only 3.7% of the phosphorylase kinase activity found in extracts of the control strain (C57BL), incubation of I strain hemidiaphragms in Krebs-Ringer bicarbonate buffer with either isoproterenol or epinephrine resulted in a stimulation of the rate of glycogenolysis. In C57BL diaphragms, the EC50 values for isoproterenol and epinephrine were 2 and 14 nM, respectively. With I strain diaphragms, dl-isoproterenol or l-epinephrine stimulated glycogenolysis as a linear function of the log of the drug concentration with no apparent plateau of response up to concentrations of 30 to 40 mugM. For each 10-fold increase in drug concentration, isoproterenol and epinephrine stimulated glycogenolysis in I strain muscles an additional 0.37 to 0.42 mg/g/hr, a slope in the concentration-response relationship of 0.17 and 0.37, respectively, of that measured in C57BL diaphragms at concentrations around the EC50. The highest glycogenolytic response measured in I strain hemidiaphragms (at 40 mugM isoproterenol) was 80% of the maximal catecholamine-stimulated glycogenolysis in C57BL diaphragms. Both 4 nM and 4 mugM isoproterenol, in a concentration-dependent manner, stimulated phosphorylase b to a conversion in I and C57BL diaphragms and increased cyclic adenosine 3':5'-monophosphate (cyclic AMP) concentrations. The glycogenolytic response to 10.1 nM dl-isoproterenol in both I and C57BL diaphragms was blocked by 34 nM l-propranolol but not by 34 nM d-propranolol. The response to 4 mugM isoproterenol was enhanced by the cyclic nucleotide phosphodiesterase inhibitors papaverine (27 mugM) or dl-4-(3-butoxy-4-methoxybenzyl)-2-imidazolidinone (Ro 20-1724, 3 mugM). From the results of these studies, we conclude: 1) Catecholamines stimulate glycogenolysis in skeletal muscle of I mice, as in C57BL mice, by interacting with the beta adrenergic receptor, thereby increasing tissue cyclic AMP concentrations and stimulating phosphorylase b to a conversion. 2) alternative hypotheses for the mechanism of the catecholamine-stimulated decrease in glycogen concentration in I skeletal muscle-inhibition of glycogen synthesis, hyposia and 5'-AMP stimulation of phosphorylase b activity-have been ruled out. 3) the activity of the mutant phosphorylase kinase, although it is only 3.7% of that in extracts of C57BL muscle, is sufficient to produce phosphorylase b to a conversion and thereby account for the glycogenolytic response of I strain muscle to catecholamines.