The Synergy Between PAV and AdaBoost

Schapire and Singer's improved version of AdaBoost for handling weak hypotheses with confidence rated predictions represents an important advance in the theory and practice of boosting. Its success results from a more efficient use of information in weak hypotheses during updating. Instead of simple binary voting a weak hypothesis is allowed to vote for or against a classification with a variable strength or confidence. The Pool Adjacent Violators (PAV) algorithm is a method for converting a score into a probability. We show how PAV may be applied to a weak hypothesis to yield a new weak hypothesis which is in a sense an ideal confidence rated prediction and that this leads to an optimal updating for AdaBoost. The result is a new algorithm which we term PAV-AdaBoost. We give several examples illustrating problems for which this new algorithm provides advantages in performance. Editor: Robert Schapire     

Relationship between surface μ‐roughness and interface slurry particle spatial distribution during glass polishing

During optical glass polishing, a number of interactions between the workpiece (i.e., glass), polishing slurry, and pad can influence the resulting workpiece roughness at different spatial scale lengths. In our previous studies, the particle size distribution of the slurry, the pad topography, and the amount of material removed by a single particle on the workpiece were shown to strongly correlate with roughness at AFM scale lengths (nm‐μm) and weakly at μ‐roughness scale lengths (μm‐mm). In this study, the polishing slurry pH and the generation of glass removal products are shown to influence the slurry particle spatial and height distribution at the polishing interface and the resulting μ‐roughness of the glass workpiece. A series of fused silica and phosphate glass samples were polished with various ceria and colloidal silica slurries over a range of slurry pH, and the resulting AFM roughness and μ‐roughness were measured. The AFM roughness was largely invariant with pH, suggesting that the removal function of a single particle is unchanged with pH. However, the μ‐roughness changed significantly, increasing linearly with pH for phosphate glass and having a maximum at an intermediate pH for fused silica. In addition, the spatial and height distribution of slurry particles on the pad (as measured by laser confocal microscopy) was determined to be distinctly different at low and high pH during phosphate glass polishing. Also, the zeta potential as a function of pH was measured for the workpiece, slurry, and pad with and without surrogate glass products (K₃PO₄ for phosphate glass and Si(OH)₄ for silica) to assess the role of interfacial charge during polishing. The addition of K₃PO₄ significantly raised the zeta potential, whereas addition of Si(OH)₄ had little effect on the zeta potential. An electrostatic DLVO three‐body force model, using the measured zeta potentials, was used to calculate the particle–particle, particle–workpiece, and particle–pad attractive and repulsive forces as a function of pH and the incorporation of glass products at the interface. The model predicted an increase in particle–pad attraction with an increase in pH and phosphate glass products consistent with the measured slurry distribution on the pads during phosphate glass polishing. Finally, a slurry “island” distribution gap (IDG) model has been formulated which utilizes the measured interface slurry distributions and a load balance to determine the interface gap, the contact area fraction, and the load on each slurry “island”. The IDG model was then used to simulate the workpiece surface topography and μ‐roughness; the results show an increase in roughness with pH similar to that observed experimentally.     

Electron microscopic characterization of Cr/silica catalyst for ethylene polymerization

Ethylene polymerization was carried out over both porous and non-porous 5 wt% Cr/silica catalysts in a slurry reactor. The polymerization was stopped at selected times to obtain samples for SEM and TEM characterization. Despite the different physical characteristics of the two silica-supported catalysts and their different behavior in the early stages of reaction, high resolution SEM micrographs (taken after runs of longer duration) revealed similar, fibrous and very porous polymer layers on both. This accessibility of the ethylene enables transport of monomer to the active sites at the very high reaction rates.     

A further investigation of the pretest-treatment interaction effect.

What effect does a pretest have on a study of an attitude-treatment when the topic is of great concern to the Ss? Segregation and vivisection were the topics in comparing pretest-treatment interaction. 224 college students plus 100 others were used as Ss. "Apparently, an attitudinal pretest has no effect on the reception of a succeeding persuasive communication within the limits of involvement of S with the topical continuum represented by vivisection… at one point and ethnic prejudice at another." (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Three theoretical interpretations of order effects in persuasive communications.

The "set," "reinforcement," and "sensory variation" hypotheses are reviewed in terms of their ability to predict the results of order effects in the persuasive situation where opposed arguments on the same topic are utilized. The set hypothesis is most successful when the communicative materials presented are unfamiliar to the S. The sensory variation hypothesis is most successful when the topics of the cummunications are concerned with familiar social issues. (24 ref.) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Book reviews

Urban Planning in a Multicultural Society

Michael A. Burayidi (ed.) Praeger, Westport, Conn., 2000 ISBN 0275961257 (hardback) 280 pp.; index RRP $123.85

The Community Planning Handbook: How People Can Shape Their Cities, Towns and Villages

Nick Wates (ed.) Earthscan, London, 2000 ISBN 1853836540 (paperback) 230 pp. RRP $42.85

City Against Suburb: The Culture Wars in an American Metropolis

Joseph A. Rodriguez Praeger, Westport, Conn., 1999 ISBN 027596406X (hardback) 144 pp.; bibliographic essay RRP $96.97

Community Development in the Market Economy

Jeremy McArdle Vista, Melbourne, 1999 ISBN 0958649677 (paperback) 117 pp.; bibliography RRP $16.50

Ethnicity and Housing: Accommodating Difference

Frederick W. Boal (ed.) Ashgate, Aldershot, 2000 ISBN 1859725961 (hardcover) 318 pp.; bibliography and index RRP $101.08

Computer‐Aided Policymaking: Lessons from Strategic Planning Software

Ray Wyatt, E. & F. N. Spon, London, 1999 ISBN 0419244808 (paperback) 244 pp.; bibliography and index RRP $79.95

Making Common Ground: Public‐Private Partnership in Land for Housing.

Geoffrey Payne (ed.). Intermediate Technology Publications, London, 1999. ISBN 1853394793 (paperback). 256 pp.; bibliography and index. RRP $45.86     

Microbiology of wheat and flour milling in Australia

A survey was undertaken to determine the microbiological status of Australian wheat and the distribution of microorganisms in the flour milling fractions and end products. A total of 650 milling process and end product samples was obtained from nine flour mills located in New South Wales (4), Queensland (2), Victoria (2) and Western Australia (1) during the 1997-1998 and 1998-1999 wheat seasons. Most frequent (modal) counts in wheat and flour were, respectively, as follows: aerobic mesophilic plate count, 10(5) and 10(2) colony forming units/gram (cfu/g); coliforms, 10 and 1 most probable number/gram (MPN/g); Bacillus spp., 10(4) and 10(2) cfu/g; B. cereus, 1 and 0.1 MPN/g; mesophilic aerobic spores, 10 and 1 cfu/g; aerobic thermophiles, both 10 cfu/g; yeasts, 10(3) and 10(2) cfu/g, and moulds, 10(3) and 10(2) cfu/g. Bacillus spp., coliforms, yeasts and moulds were the most frequently detected microorganisms throughout the survey. The most common moulds isolated were Aspergillus, Penicillium, Cladosporium and Eurotium spp. Environmental serovars of Salmonella were isolated from two samples. Escherichia coli and B. cereus were present at very low levels, a majority of positive samples being at the minimum level of detection (3 and 0.3 MPN/g, respectively). As wheat grain layers are separated, surface-adhering contaminants are concentrated in end product bran, wheat germ and pollard, which comprise the outer layers of the grain. Consequently, the inner endosperm fraction contains lower microbial counts, and flour is the cleanest end product of the milling process. Higher microbiological counts midstream in the milling process indicate that equipment contamination may contribute to microbiological contamination; however, the microbiological quality of incoming wheat has a strong influence on the ultimate quality of milling end products.     

Study of the dispersion of natural gas issuing from compressor stations through silencers with upper cover

The aim of the present study is the simulation of the dispersion of natural gas issuing from the silencer of compressor stations during vent operations. The objective is to analyze the dispersion of the gas emitted under different conditions of mass flow rate at the exit and ambient cross-flow velocity. We have considered a silencer with an upper cover to protect it from the rain and the fall of objects. The influence of the upper cover of the silencer on the dispersion of natural gas has also been studied, and non-dimensional approaches of the model have been proposed to simplify the problem. Seven different cases have been solved, using two models: a 3D model based on the commercial code FLUENT, and a simplified quasi-one-dimensional model. The results obtained in both cases have been compared, and the range of validity of the one-dimensional model in non-dimensional form has been discussed.     

Ion Diffusion Study in the Oxide Layers Due to Oxidation of AISI 439 Ferritic Stainless Steel

Chromium diffusion coefficients were determined in thermally grown chromia layers on the AISI 439 ferritic stainless steel oxidized at 750, 800, 850 and 900 °C in synthetic air. The stable isotope ⁵⁴Cr was used as chromium tracer and the diffusion profiles were established by SIMS analysis. The chromium ion bulk diffusion coefficients evaluated numerically from experimentation were around five orders of magnitude smaller than the chromium ion diffusion coefficients at grain-boundary regions. The values of the chromium ion diffusivities in chromia grown on the AISI steel were lower than the corresponding oxygen ion diffusivities obtained in a previous study. Comparison of experimental and calculated parabolic oxidation constants, for the oxidation of the AISI 439 steel, showed that the chromia scale growth mechanism is controlled not only by inward oxygen ion diffusion from atmosphere, but also by outward chromium diffusion from the metallic substrate. However, the role of the oxygen ion diffusion appears to be more important than that of the chromium ion diffusion.