Evaluating Classification Model Against Bayes Error Rate.

For a classification task, we usually select an appropriate classifier via model selection. How to evaluate whether the chosen classifier is optimal? One can answer this question via Bayes error rate (BER). Unfortunately, estimating BER is a fundamental conundrum. Most existing BER estimators focus on giving the upper and lower bounds of the BER. However, evaluating whether the selected classifier is optimal based on these bounds is hard. In this article, we aim to learn the exact BER instead of bounds on BER. The core of our method is to transform the BER calculation problem into a noise recognition problem. Specifically, we define a type of noise called Bayes noise and prove that the proportion of Bayes noisy samples in a data set is statistically consistent with the BER of the data set. To recognize the Bayes noisy samples, we present a method consisting of two parts: selecting reliable samples based on percolation theory and then employing a label propagation algorithm to recognize the Bayes noisy samples based on the selected reliable samples. The superiority of the proposed method compared to the existing BER estimators is verified on extensive synthetic, benchmark, and image data sets.

Real-time batch inspection system for surface defects on circular optical filters.

Optical filters, one of the essential parts of many optical instruments, are used to select a specific radiation band of optical devices. There are specifications for the surface quality of the optical filter in order to ensure the instrument's regular operation. The traditional machine learning techniques for examining the optical filter surface quality mentioned in the current studies primarily rely on the manual extraction of feature data, which restricts their ability to detect optical filter surfaces with multiple defects. In order to solve the problems of low detection efficiency and poor detection accuracy caused by defects too minor and too numerous types of defects, this paper proposes a real-time batch optical filter surface quality inspection method based on deep learning and image processing techniques. The first part proposes an optical filter surface defect detection and identification method for seven typical defects. A deep learning model is trained for defect detection and recognition by constructing a dataset. The second part uses image processing techniques to locate the accurate position of the defect, determine whether the defect is located within the effective aperture, and analyze the critical eigenvalue data of the defect. The experimental results show that the method improves productivity and product quality and reduces the manual workload by 90%. The proposed model and method also compare the results of surface defect detection with the actual measurement data in the field, verifying that the method has good recognition accuracy while improving efficiency.

Cerium Addition Improved the Dry Sliding Wear Resistance of Surface Welding AZ91 Alloy.

In this study, the effects of cerium (Ce) addition on the friction and wear properties of surface welding AZ91 magnesium alloys were evaluated by pin-on-disk dry sliding friction and wear tests at normal temperature. The results show that both the friction coefficient and wear rate of surfacing magnesium alloys decreased with the decrease in load and increase in sliding speed. The surfacing AZ91 alloy with 1.5% Ce had the lowest friction coefficient and wear rate. The alloy without Ce had the worst wear resistance, mainly because it contained a lot of irregularly shaped and coarse ß-Mg17Al12 phases. During friction, the ß phase readily caused stress concentration and thus formed cracks at the interface between ß phase and α-Mg matrix. The addition of Ce reduced the size and amount of Mg17Al12, while generating Al4Ce phase with a higher thermal stability. The Al-Ce phase could hinder the grain-boundary sliding and migration and reduced the degree of plastic deformation of subsurface metal. Scanning electron microscopy observation showed that the surfacing AZ91 alloy with 1.5% Ce had a total of four types of wear mechanism: abrasion, oxidation, and severe plastic deformation were the primary mechanisms; delamination was the secondary mechanism.

[Primary study on contents of sulfate reducing bacteria (SRB) and organic matter from intertidal zone at Chongming Dongtan].

Collected soil samples from different tidal flats and elevation in Chongming Dongtan wetland, then conducted sulfate-reducing bacteria (SRB) based on MPN method, determined organic matter content and calculated SO4(2-)/Cl- molar ratio, for the research on the distribution of SRB, relevance to soil organic matter content as well as influence of plant rhizosphere environment on SRB growth. The results show the distribution of SRB is ranked as middle flat > climax flat > bald flat. The same tidal flats at different depths, the SRB levels are shown as 51-52 cm > 21-22 cm > 81-82 cm, therefore 51-52 cm soil depth of Dongtan wetland is the suitable area for SRB to grow. However, in different tidal and depth, the distribution of organic matter content presents climax flat > middle flat > bald flat. From 21-51 cm, as the depth increasing, the organic matter content decreases while the amount of SRB significantly increasing, which indicates SRB utilizes the soil organic matter to carry out reduction reaction. The SO4(2-)/Cl- molar ratios of all soil samples are less than 0.05, indicating that SRB are actively engaged in sulfate reduction. The concentration of SRB in reed rhizosphere soil is the highest, showing that Phragmites australis rhizosphere environment in Dongtan wetland could enhance SRB growth, while the number of SRB in Spartina alterniflora rhizosphere environment is relatively lower than the non-rhizosphere environment, indicating that the rhizosphere effect has different effects on SRB in Dongtan tidal flats.

[Quantitative simulation on the vertical distribution of soil organic matters in mountainous soil profiles in the subtropical area, south China].

Quantitative descriptions of soil organic matter (SOM) dynamics, i.e., their distribution, turnover and movement, are essential for the running of the simulation of terrestrial ecosystem organic matter models. In this study, based on utilizing SOM diffusion-translation-decomposition model, two soil profiles were selected in different vegetation zones at Dinghushan Mountain for quantitative studies on SOM dynamics and their controlling factors. SOM were divided into three kinds of compartments: rapid compartment with turnover rate of 0.1-1.yr-1, slow compartment with turnover rate of 0.002-0.02.yr-1, and stable compartment with turnover rate of 0.0001-0.001.yr-1. The numerical results suggested that SOM distribution in soil profile in subtropical mountainous areas of south China obeyed the law of diffusion motion, translation motion and decomposition. The turnover rate of SOM rapid compartment was 0.483.yr-1 in the forest vegetation zone, and was 0.694.yr-1 in the shrub vegetation zone. The turnover rates of SOM slow compartment in the two kinds of vegetation zones were both 0.02.yr-1, and the turnover rates of SOM stable compartment in the two kinds of vegetation zones were both 0.001.yr-1. SOM diffusion rate and translation rate for the forest vegetation zone was 4 cm2.yr-1 and 0.2 mm.yr-1, respectively, and the two rates of the shrub vegetation zone were 1 cm2.yr-1 and 0.5 mm.yr-1, respectively. The obvious discrepancy between numerical values and measuring values for SOM content occurred in the 0-10 cm sections of the profiles, which might be due to the fact that the upper sections were at the interface between lithosphere and atmosphere, and were influenced directly by changes of climatic and environmental factors. The two kinds of values for SOM content were identical below the upper section of the profiles, and it indicated stable pedogenesis environments. Diffusion motion had obvious influences on SOM vertical distribution, and translation motion had clear impacts on SOM distribution only in the upper 0-10 cm section. Comparison analysis suggested that SOM dynamics were controlled mainly by soil profile qualities such as SOM content, clay content, soil fabric, void types and their developments, soil fauna and microorganism activities, etc. With the increasing of primary production of aboveground vegetation, the turnover rate of SOM rapid compartment decreased and SOM content increased, which provided scientific basis for increasing soil carbon sink through anthropogenic effects.