Adaptive Sampling Methods for Scaling Up Knowledge Discovery Algorithms

Scalability is a key requirement for any KDD and data mining algorithm, and one of the biggest research challenges is to develop methods that allow to use large amounts of data. One possible approach for dealing with huge amounts of data is to take a random sample and do data mining on it, since for many data mining applications approximate answers are acceptable. However, as argued by several researchers, random sampling is difficult to use due to the difficulty of determining an appropriate sample size. In this paper, we take a sequential sampling approach for solving this difficulty, and propose an adaptive sampling method that solves a general problem covering many actual problems arising in applications of discovery science. An algorithm following this method obtains examples sequentially in an on-line fashion, and it determines from the obtained examples whether it has already seen a large enough number of examples. Thus, sample size is not fixed a priori; instead, it adaptively depends on the situation. Due to this adaptiveness, if we are not in a worst case situation as fortunately happens in many practical applications, then we can solve the problem with a number of examples much smaller than required in the worst case. We prove the correctness of our method and estimates its efficiency theoretically. For illustrating its usefulness, we consider one concrete task requiring sampling, provide an algorithm based on our method, and show its efficiency experimentally.     

SIR Weighting Combining Diversity Using Matched Filter for Personal Communication Systems

We propose a novel SIR weighting postdetection combining diversity scheme with a new accurate SIR estimation method. The SIR is estimated and used as the weighting factor to compensate severe cochannel interference, one of the most important issues for PCS in terms of frequency utilization. Theimprovement offered by the proposal depends on SIR estimation accuracy.The SIR is, in this paper, estimated by a matched filter where theauto-correlation between received signal and unique word is calculated. Computer simulationsconfirm that the SIR of each diversity branch can be estimated easily andaccuratelyby the proposed SIR estimation method. The proposed diversity scheme achievesaperformance very close to that of ideal SINR weighting diversity underRayleighfading with severe cochannel interference. When average SIR = 10 dB and thenumber of branches(L)=4, the proposed diversity scheme lowers the requiredE_b/N₀ by 5 dB at BER = 1×10^-3compared to conventional maximal ratio combining diversity. This paper alsopresentsthe unique word length required to realize adequate performance, i.e.,robustnessagainst high-pitch Rayleigh fading.     

The metabolism and distribution of docosapentaenoic acid (n−6) in rats and rat hepatocytes

In this study, a new marine oil that contains 45% docosahexaenoic acid (DHA, 22∶6n−3) and 13% docosapentaenoic acid (DPA, 22∶5n−6) was administered to rats. The metabolism and distribution of DPA in rats was investigated. In experiment 1, the effects of DHA and n−6 fatty acids (linoleic acid, I A; arachidonic acid, AA; and DPA) on AA contents were investigated in vivo. LA group: LA 25%, DHA 30%; LA-DPA group: LA 15%, DPA 10%, DHA 35%; LA-AA-DPA group: LA 10%, AA 5%, DPA 10%, DHA 35% were administered to rats for 4 wk. In the liver, the AA content in the LA-DPA and LA-AA-DPA groups was significantly higher than in the LA group. The decreased AA contents in the LA group might be caused by DHA administration. Although DHA also was administered in the LA-DPA and LA-AA-DPA groups, the AA contents in these two groups did not decrease. These results suggested that DPA retroconverted to AA, blunting the decrease in AA content caused by DHA administration. To conduct a detailed investigation on DPA metabolism and its relation with AA and DHA, rat hepatocytes were cultured with pruified DPA and DHA for 24 h. We discovered the retroconversion of DPA to AA occurred only when AA content was decreased by a high DHA administration; it did not occur when AA content was maintained at a normal level.     

Enhancement of the thermoelectric figure of merit in p- and n-type Cu/Bi-Te/Cu composites

The resultant thermoelectric properties of welded Cu/Bi-Te/Cu composites were measured at 298 K as a function of relative thickness x of Bi-Te compound by changing the interval s between two thermocouples and compared with those calculated as a function of x by treating it as an electrical and thermal circuit. These composites were prepared by welding with eutectic solder of Pb-Sn, after one end surface of the as-grown p- and n-type Bi-Te ingots were plated with Ni. It was found that the observed ZT of composites has a local maximum at an optimum x even when s was changed, as in the case of Cu/Bi-Sb/Cu and Ni/Bi-Sb/Ni composites with various relative thicknesses. Appearance of a local maximum in ZT is owing to the barrier thermo-emf generated by a sharp temperature drop at the interface between Bi-Te compound and copper. It may be caused by the separation of non-equilibrium carriers at the interface between them. The observed maximum ZT values at 298 K of the p- and n-type composites reached surprisingly great values of 1.53 and 1.66 at x=0.98, which correspond to about twice as large as those of commercially utilized Bi-Te compounds. This enhancement of ZT is available for generators, but may be not utilizable as a Peltier module. The composite materials were thus found to be utilizable as useful means of further increase in ZT of macroscopic bulk materials.     

Reliable indirect competitive ELISA used for a survey of ochratoxin A in green coffee from the North of Paraná State, Brazil.

The performance of an indirect competitive enzyme-linked immunosorbent assay (ic-ELISA) based on a monoclonal antibody (mAb) for ochratoxin A (OTA) detection was evaluated in a comparative study with high-performance liquid chromatography (HPLC) analysis using 68 freshly harvested coffee samples from the North of Paraná State, Brazil. The anti-OTA mAb showed high specificity and low cross-reactivity with OTA analogues (OTB and OTalpha), but cross-reacted with OTC. This ic-ELISA showed a detection limit of 3.75 ngg-1 sample, when compared to 0.80 ngg-1 by HPLC, with an ic-ELISA/HPLC correlation coefficient of 0.90. As regards OTA analysis of these coffee samples, natural contamination was detected in 10 samples (14.7%) by both methods, where the ic-ELISA values (range 3.9-7.3 ngg-1) were 1.1 to 1.6-fold higher than HPLC data (2.7-4.7 ngg-1). Five samples (7.4%) were OTA positive (range 0.84-1.30 ngg-1) only by HPLC assay, probably due to the higher detection limit reached by ic-ELISA. OTA was undetectable in 53 samples (77.9%) by both methods, while all positive samples (range 0.84-7.30 ngg-1) showed OTA levels lower than 8 ngg-1 (maximum limit recommended by the European Union). The matrix interference of green coffee was minimized by dilution of sample extracts before carrying out the ELISA assay. This mAb-based ic-ELISA can be effectively applied for OTA screening in coffee, because it is simple, sensitive and sample preparation is easy.     

Surface structures of synthetic diamonds

Diamond crystals of about 2 mm in size were grown by the temperature gradient method using Ni, Ni-Fe alloy and Fe as solvent metals. Several characteristic patterns were seen on the surfaces of the grown diamonds and were found to correspond to the texture of the quenched solvent metals. A vague vein-like pattern appeared when Ni was used. The surface patterns became finer with the increase of the content of Fe in Ni-Fe alloy, and when 70 wt% Fe-30 wt% Ni alloy was used, very fine dendritic patterns were seen on all of the faces of diamond in accordance with the symmetries of the lattice. The patterns are formed by the precipitation of carbon atoms in the space of the dendrites of the solvent metals during the quenching process. Lamellar patterns and etch pits were seen when Fe was used. The patterns are formed when Fe dissolves the diamond surfaces to produce Fe₃C during the quenching process.     

The relationship between ductility and grain-boundary precipitates in an Al-6.0% Zn-2.6% Mg alloy

The relationship between ductility and material parameters such as grain-boundary (GB) precipitate size and dimple size in an Al-6.0% Zn-2.6% Mg alloy has been studied. Under the condition of a given GB precipitate size, the ductility increases with decreasing number of GB precipitates and dimples per unit area. The critical GB precipitate sizes acting as the nuclei for void formation are deduced.     

Isolation and characterization of a novel gene encoding alpha-L-arabinofuranosidase from Aspergillus oryzae

We cloned and characterized a novel gene (abfA) encoding alpha-L-arabinofuranosidase (alpha-L-AFase) from Aspergillus oryzae. One clone homologous to the alpha-L-AFase gene of Thermotoga maritima was found in an expressed sequence tag (EST) library of A. oryzae and a corresponding gene was isolated. Molecular analysis showed that the abfA gene carried six exons interrupted by five introns and had an open reading frame encoding 481 amino acid residues. The amino acid sequence similarity at active sites to the alpha-L-AFases from other organisms indicated that the alpha-L-AFase encoded by abfA was classified as a family 51 glycoside hydrolase. When the abfA was overexpressed in the homologous hyperexpression system of A. oryzae, a large amount of alpha-L-AFase was produced as intracellular protein. The apparent molecular mass of the purified enzyme was estimated to be 228,000 by gel filtration and that of its subunit as 55,000 by SDS-PAGE, suggesting that the enzyme is a tetramer. The enzyme hydrolyzed p-nitrophenyl-alpha-L-arabinofuranoside but not other p-nitrophenyl glycosides. These results demonstrated that the abfA gene encodes a functional alpha-L-AFase.     

Effect of lattice constant of zinc oxide on antibacterial characteristics

Zinc oxide powders were heated in different atmospheres at 800 and 1400 degrees C, of which the characterization and the antibacterial activity were studied by X-ray diffractometry and the measurement of the change in electrical conductivity with bacterial growth. The diffraction peaks corresponding to zinc oxide with hexagonal type structure were detected in all samples, which shifted in low-angle side with the increase in the oxidizability of atmosphere during heat-treatment. From the results of calculating lattice constants, a0 and c0, it was found that the value of c0 in hexagonal structure increased with the increase in the oxidizability of atmosphere. On the samples heated at 1400 degrees C, the changes of the c0 value were less than those at 800 degrees C. However, no change of the a0 value showed, irrespective of atmosphere and temperature. Hydrogen peroxide that contributes to the occurrence of antibacterial activity was found to generate from all samples, and the generation amount increased with the increase of c0 value; incidently the amount in the samples heated at 1400 degrees C was less than that at 800 degrees C. The antibacterial activity of zinc oxide increased with the increase of c0 value; that is, it was found that the value of c0 in crystal structure affected the antibacterial activity of zinc oxide.