A study of SVM using a combination of the online learning method and the midpoint-validation method

The support vector machine (SVM) is known as one of the most influential and powerful tools for solving classification and regression problems, but the original SVM does not have an online learning technique. Therefore, many researchers have introduced online learning techniques to the SVM. In a previous article, we proposed an unsupervised online learning method using the technique of the self-organized map for the SVM. In another article, we proposed the midpoint validation method for an improved SVM. We test the performance of the SVM using a combination of the two techniques in this article. In addition, we compare its performance with the original hard-margin SVM, the soft-margin SVM, and the k-NN method, and also experiment with our proposed method on surface electromyogram recognition problems with changes in the position of the electrode. These experiments showed that our proposed method gave a better performance than the other SVMs and corresponded to the changing data.     

Demonstration of loss reduction effect of 66 kV‐classed 30 MVA three phase hybrid‐core transformer by trial manufacture

To realize larger‐capacity “Hybrid‐core transformer: HBT”, this research has assembled a 30‐MVA three‐phase trial HBT and evaluated the loss performances. The hybrid‐core consisting of the wound amorphous‐ and the stacked silicon steel‐cores, is expected having advantages of lower iron loss of the iron‐based amorphous material and the higher mechanical strength and saturation magnetic flux density of the silicon steels simultaneously. Three‐dimensional finite‐element method analysis revealed that the silicon steel core prevented over‐saturation of the amorphous core. The hybrid configuration enabled the design to have an approximately 10% higher magnetic flux density than configurations with only an amorphous core. We then assembled a prototype 30‐MVA three‐phase HBT designed on the basis of our investigation results. Operation tests demonstrated that the HBT has 62%‐reduced iron loss and 0.10%‐higher 50%‐loaded power efficiency from the conventional silicon steel core transformer.     

Value-utilized taint propagation: toward precise detection of apps’ information flows across Android API calls

International Journal of Information Security - Android security researchers utilize taint analysis to uncover apps’ bugs and policy-violating behaviors. However, the investigations are...     

Purification of Branched Dextrin from Nägeli Amylodextrin by Ethanol Precipitation and Characterization of Its Aggregation Property in Methanol-Water

Ethanol precipitation process for purification of branched dextrin (BD) in Nägeli amylodextrin from waxy rice starch was developed. Temperature and ethanol concentration for precipitation were main parameters affecting the recovery and purity of BD, and the purification condition at 4 °C and 10 % (v/v) ethanol in water was adopted. After four-time precipitation, the BD recovery was 34.6 %, whereas the purity improved from 78.5 % at the initial to 94.5 % at the four-time purified BD (BD4). BD4 mainly showed a chain length distribution between 18 to 35 with a mode length of 25, which shifted after enzymatic debranching with isoamylase to that between 9 and 20 with a mode length of 14. Each purified BD was solubilized in water, and each solution was mixed with methanol-water at 25 °C to a final methanol concentration of 16 M. The flakes of BD precipitated with 16 M methanol exhibited an A-type crystal structure by an X-ray diffraction analysis, and the speed generation of white flakes in 16 M methanol dramatically increased as the purification time increased. The effect of addition of highly branched cyclic dextrin (HBCD) or sodium tetraborate on BD aggregation in 16 M methanol was also investigated, where the former retarded aggregation but the latter had no effect on the velocity. Thus, the purified BD enables rapid characterization of aggregation of double helix structures of A-type crystal structure, and screening of compounds which could affect the phenomena for prediction of potentials in starch modification as well.     

Self-crack healing ability and strength recovery in ytterbium disilicate/silicon carbide nanocomposites

The SiO₂ volatility and SiC recession in hydrocarbon-rich combustion atmosphere is the main problem of SiC-reinforced ceramics when they are utilized as coating materials for gas turbine blades. The composite of Yb₂Si₂O₇-Yb₂SiO₅/SiC is expected as a self-crack healing material that can avoid this problem because the monosilicate can react with the SiO₂ glass to form disilicate and hence further reinforce the composite. In this study, the composites, fabricated with various morphology of SiC nanofillers, were precracked and then heat-treated in an oxidizing environment to activate their crack healing behavior. The healing effect as a function of filler morphology and annealing time was investigated. Healing mechanism and future applications of these composites were also discussed.     

Characterization of a GH Family 20 Exo-β-N-acetylhexosaminidase with Antifungal Activity from Streptomyces avermitilis

We characterized SaHEX, which is a glycoside hydrolase (GH) family 20 exo-β-N-acetylhexosaminidase found in Streptomyces avermitilis. SaHEX exolytically hydrolyzed chitin oligosaccharides from their non-reducing ends, and yielded N-acetylglucosamine (GlcNAc) as the end product. According to the initial rate of substrate hydrolysis, the rates of (GlcNAc)₃ and (GlcNAc)₅ hydrolysis were greater than the rates for the other oligosaccharides. The enzyme exhibited antifungal activity against Aspergillus niger, which was probably due to hydrolytic activity with regard to chitin in the hyphal tips. Therefore, SaHEX has potential for use in GlcNAc production and food preservation.