Search

This paper examines a number of stochastic computational elements employed in artificial neural networks, several of which are introduced for the first time, together with an analysis of their operation. We briefly include multiplication, squaring, addition, subtraction, and division circuits in both unipolar and bipolar formats, the principles of which are well-known, at least for unipolar signals. We have introduced several modifications to improve the speed of the division operation. The primary contribution of this paper, however, is in introducing several state machine-based computational elements for performing sigmoid nonlinearity mappings, linear gain, and exponentiation functions. We also describe an efficient method for the generation of, and conversion between, stochastic and deterministic binary signals. The validity of the present approach is demonstrated in a companion paper through a sample application, the recognition of noisy optical characters using soft competitive learning. Network generalization capabilities of the stochastic network maintain a squared error within 10 percent of that of a floating-point implementation for a wide range of noise levels. While the accuracy of stochastic computation may not compare favorably with more conventional binary radix-based computation, the low circuit area, power, and speed characteristics may, in certain situations, make them attractive for VLSI implementation of artificial neural networks.

For pt. I see ibid., p.891-905. An investigation has been made into the use of stochastic arithmetic to implement an artificial neural network solution to a typical pattern recognition application. Optical character recognition is performed on very noisy characters in the E-13B MICR font. The artificial neural network is composed of two layers, the first layer being a set of soft competitive learning subnetworks and the second a set of fully connected linear output neurons. The observed number of clock cycles in the stochastic case represents an order of magnitude improvement over the floating-point implementation assuming clock frequency parity. Network generalization capabilities were also compared based on the network squared error as a function of the amount of noise added to the input patterns. The stochastic network maintains a squared error within 10 percent of that of the floating-point implementation for a wide range of noise levels.

The immunomodulatory potential of dietary chromium (Cr) has been demonstrated in cattle. Specific Cr-attributed health and production benefits, in both beef and dairy cattle, are postulated to arise in part from the direct and/or indirect consequences of this immunomodulatory potential and from the long-recognized effects of Cr on insulin and glucose metabolism. Recent studies in dairy heifers evaluated: (1) peripheral tissue responsiveness to insulin following feed supplementation with varying concentrations of Cr-yeast, and (2) blood lymphocyte (BL) insulin receptor expression and binding affinity following dietary supplementation with Cr-chelate. These findings suggest that Cr potentiation of insulin action may be mediated primarily by postreceptor pathways. In addition, in vitro studies examined the effect of Cr-supplemented culture medium on the quantitative expression of several cytokine mRNAs in concanavalin A (ConA)-stimulated bovine BLs. These findings support a previously reported Cr-mediated decrease in interleukin-2 (IL-2) activity in the culture medium of ConA-stimulated BLs from Cr-supplemented cows. They further suggest a role for immunoregulatory cytokines, particularly IL-2, as a mechanism through which Cr exerts its effects upon the bovine immune response. Additionally, the differential inhibition of bacterial growth rates by Cr in vitro, simulating possible effects upon anaerobic rumen bacteria populations, suggests a potential indirect mechanism by which dietary Cr may influence ruminant health and performance. The next generation of studies elucidating the cellular and molecular mechanism of Cr immunomodulatory effects will benefit greatly from a reliable, chemically defined source of Cr bioplexes and from the experience with Cr in ruminants. J. Trace Elem. Exp. Med. 12:131–140, 1999. © 1999 Wiley-Liss, Inc.

The transfection efficiencies of a panel of eight uniquely different lipid reagents has been evaluated with two other commercially available lipids for use in transfecting a diversity of eukaryotic cell lines. The PerFect™ lipids are available individually or together in an optimization panel format that can be tested in any given cell line, enabling one to evaluate the optimal lipid for transfecting each individual cell line. Our results demonstrate that no single lipid is optimal for plasmid transfection over a broad range of cell types, thus emphasizing the need for multiple unique lipid reagents and a simple format for testing their transfection efficiency on a given cell type.

A novel circuit for I/sub ddq/ monitoring is presented. The circuit is based on current mode signaling and is intended for use in a built-in test mode. A test chip has been fabricated in 1.2 micron CMOS. Preliminary simulation result are very promising and demonstrate that the basic technique may be quite practical.

We have adapted a non-radioactive technique to detect localized mRNAs in whole-mount Xenopus embryos. Synthetic antisense RNA transcribed in the presence of digoxygenin-UTP is used as a probe and is detected via an anti-digoxygenin antibody. We show that localized mRNAs can be detected from late gastrula to tadpole stages and that high as well as low abundance RNAs can be detected. The method was tested on muscle actin and α-globin RNAs, whose localization has previously been characterized. In addition, we used the method to determine the distribution of XA-1 RNA, an anterior ectoderm-specific RNA, which we show is expressed in the periphery of the cement gland as well as in the region of the hatching gland. The sequence of an XA-1 cDNA predicts a protein rich in proline and histidine.

The composition of surfactant from the lungs of babies dying from the sudden death syndrome (SIDS) differs from that of babies and infants dying of other causes. Surfactant from SIDS babies contained significantly less phospholipid and dipalmitoylphosphatidylcholine than does surfactant from mature babies dying of other causes. The amounts of phospholipid components were similar in surfactant from SIDS infants and those with hyaline disease, but SIDS surfactant contained more phosphatidylglycerol and less dipalmitoylphosphatidylcholine. These results implicate abnormal surfactant in the pathogenesis of SIDS.

Lung surfactant was obtained by postmortem lavage from: (A) premature babies: 34 dying acutely within 2 days of birth from Hyaline Membrane Disease (HMD), 20 dying several days after birth with HMD and its consequences, 8 dying from causes other than HMD; (B) mature babies: 24 dying stillborn, 15 dying soon after birth and 16 dying between 2 weeks and 1 year of age with minimal lung pathology. The phospholipid composition of the surfactant was analysed. Compared to the surfactant of babies dying acutely from HMD, that of the babies dying later from HMD contained significantly higher proportions of phosphatidylcholine (PC) and significantly lower proportions of sphingomyelin while that of the mature babies contained significantly higher proportions of PC and phosphatidylglycerol but significantly lower proportions of sphingomyelin and combined phosphatidylinositol and phosphatidylserine. The surfactant of premature babies dying of causes other than HMD was similar and intermediate to that of both groups of babies dying from HMD. The PC fraction composition of the surfactant of the babies dying acutely from HMD contained significantly lower proportions of the disaturated fraction than those of the babies dying later from HMD, stillborn babies or mature babies.

The lung surfactant phospholipid composition of lavage samples from 102 babies dying from Sudden Infant Death Syndrome (SIDS) (one-third with minor signs of inflammation) was compared with that of: 34 babies dying from Hyaline Membrane Disease (HMD), 15 mature babies dying soon after birth, 16 mature babies dying in the same age range as the sudden infant death syndrome cases, 13 babies dying from pneumonia and 6 from septicaemia. The surfactant of the two groups of babies dying from SIDS was identical and approximated that obtained from babies dying from HMD, pneumonia or septicaemia. Compared to that obtained from mature babies, the surfactant of babies dying from SIDS contained significantly lower proportions of phosphatidylcholine (PC) and significantly higher proportions of lyso-PC and sphingomyelin. The proportion of disaturated PC was similar to that of the surfactant of the age-matched mature babies. The surfactant composition of the babies dying from SIDS did not change appreciably after death nor vary with age at death. The surfactant phospholipid composition of postmortem samples from mature babies was similar to that of aspirates from living babies and infants and to that of bronchoalveolar lavage samples from living adults.

In rabbits fed a high cholesterol diet for 12 weeks the following were found: in serum, a decrease in lipase and no changes in phospholipase A and cholesterol esterase activities; in the aorta, decreased cholesterol esterase and acyl-CoA cholesterol acyltransferase (ACAT) activities and no changes in phospholipase A, lipase and acyl-CoA synthetase (ACS) activities; in the liver, increased phospholipase A, cholesterol esterase and ACAT activities, and decreased lipase and ACS activities. A single injection of polyunsaturated phosphatidyl choline in 4% sodium deoxycholate (EPL solution) in control and hypercholesterolaemic rabbits resulted in different effects on the enzyme activities. The stimulating, diet-independent and heparin-like effect of the drug on lipase consisted of increased activities of the enzyme in serum and liver and decreased activity in the aorta. Diet-dependent changes of the enzyme activities were as follows: in serum, decreased phospholipase A activity in control and increased phospholipase A activity in hypercholesterolaemic diet; in the aorta, decreased cholesterol esterase and elevated cholesterol ester synthesis/hydrolysis ratio in control and increased phospholipase A activity, decreased ACS activity and lowered cholesterol ester synthesis/hydrolysis ratio in hypercholesterolaemic diet; in the liver, increased phospholipase A and ACAT activities and decreased ACS activity with the ratio of cholesterol ester synthesis/hydrolysis elevated in control and this ratio unchanged in hypercholesterolaemic diet. Possible mechanisms of the enzyme activity changes are discussed.

The composition of surfactant from the lungs of neonatal puppies dying from 'fadding puppy complex' differs from that of neonates dying from other causes. Surfactant from 'fading puppies' contained significantly less phosphatidylcholine than did the surfactant from other puppies, and therefore abnormal surfactant may be implicated in fading puppy complex. Abnormal surfactant has been implicated in the pathogenesis of sudden infant death syndrome in human infants.