Visual privacy attacks and defenses in deep learning: a survey

The concerns on visual privacy have been increasingly raised along with the dramatic growth in image and video capture and sharing. Meanwhile, with the recent breakthrough in deep learning technologies, visual data can now be easily gathered and processed to infer sensitive information. Therefore, visual privacy in the context of deep learning is now an important and challenging topic. However, there has been no systematic study on this topic to date. In this survey, we discuss algorithms of visual privacy attacks and the corresponding defense mechanisms in deep learning. We analyze the privacy issues in both visual data and visual deep learning systems. We show that deep learning can be used as a powerful privacy attack tool as well as preservation techniques with great potential. We also point out the possible direction and suggestions for future work. By thoroughly investigating the relationship of visual privacy and deep learning, this article sheds insights on incorporating privacy requirements in the deep learning era.

     

Assessing students' conceptual understanding using an online three‐tier diagnostic test

When solving a mathematical problem, students who do not have sufficient conceptual understanding may perform poorly and exhibit misconceptions. This study was aimed to examine students' conceptual understanding and significant misconceptions when solving number sense‐related problems. An online three‐tier diagnostic test was administered to 125 fifth‐grade students with varied socio‐economic backgrounds in Hong Kong. Only 14.40% of the students exhibited high performance with high confidence, indicating that these students had a profound conceptual understanding of number sense. In addition, the majority of the students (66.40%) did not demonstrate number sense; these students exhibited several significant misconceptions and could solve the questions only by using a rule‐based method or guessing. Accordingly, most students performed unsatisfactorily on number sense‐related problems. This study is imperative in identifying early predictors and provides information for further compatible interventions in the teaching and learning of number sense in Hong Kong in particular and worldwide in general.     

黑色小扁豆中不同结合态酚类含量、花青素组成及其抗氧化活性研究

黑色小扁豆因含有丰富的营养素、微量营养素和植物化学物成分而受到研究者的关注。为深入了解小扁豆的营养价值,对小扁豆中不同结合形态的酚类化合物含量及其抗氧化活性(亚铁离子还原能力、DPPH自由基清除能力、ABTS自由基清除能力)进行测定。结果表明,黑色小扁豆中不同存在形式的酚类提取物含量差异显著。酯键合态酚类提取物中含有较高的酚酸含量(1.61 mg/g),碱水解结合态酚类提取物具有较高的黄酮含量(1.07 mg/g)及抗氧化活性【亚铁离子还原能力(43.80 mmol/g)、DPPH自由基清除能力(2.13 mg/g)、ABTS清除能力(3.92 mg/g)】。对黑色小扁豆中存在的花青素进行提取和分析鉴定,其含有的花青素主要为飞燕草-3-O-(2-O-β-D-吡喃葡萄糖基-α-L-吡喃阿拉伯糖苷)和矢车菊素衍生物。     

Molecularly imprinted poly(acrylonitrile‐co‐acrylic acid) matrix with sclareol

There are presented results regarding a new set of molecularly imprinted polymers (MIPs) based on acrylonitrile:acrylic acid (AN:AA) copolymer matrix. As template, it was used sclareol, an important anticancer bioactive compound, never used before for molecular imprinting. An emerging and insufficient studied MIP preparation method, namely the phase inversion, was used to prepare 0.8 mm spherical sclareol MIPs (S‐MIPs). Three AN:AA copolymers, having the initial monomer ratios 90:10, 80:20, and 70:30, were synthesized by radical copolymerization in emulsion, without emulsifier. After that, each copolymer was dissolved in the presence of the template (sclareol) in dimethylformamide. The imprinting and the morphology of these new materials were analyzed by rheology, elemental analysis, infrared spectroscopy, size exclusion chromatography, thermogravimetric analysis, differential scanning calorimetry, batch rebinding tests, and Scatchard analysis. The conclusion was that the AN:AA‐80:20 matrix proved to be the optimized solution between high rigidity (given by the AN segments) and high affinity for the template (given by AA segments), the average imprinting factor for this system being 2.67. POLYM. ENG. SCI., 54:1484–1494, 2014. © 2013 Society of Plastics Engineers     

Manufacture of three-layer wood–porcelain stone composite board reinforced with bamboo fiber

The objective of this study was to improve bending strength properties of three-layer wood–porcelain stone composite board. The focus of this study was on the effects of orientations and weight ratios of bamboo fiber in face layer on physical and mechanical properties of the board. Three types of board with different orientation of bamboo fibers in the face layer were manufactured: one in which the fibers were oriented at random orientation (R board), another in which the fibers were oriented at unidirectional orientation (U board), and a third in which the fibers were oriented at cross orientation (C board). The physical and mechanical properties of the boards were evaluated based on the Japanese Industrial Standard for Particleboards. The main results obtained were as follows: Bending strength properties of the board were strongly affected by both orientation and weight ratio of bamboo fibers. Perpendicular specimen of C board has larger bending strength properties than U board and the value become larger with increased weight ratio of bamboo fibers. Internal bond strength value decreased with increasing weight ratio of bamboo fibers. The effect of orientation and weight ratios of bamboo fiber on thickness swelling of the board was not significant.     

Study on corrosion of welded steel for LBE-cooled fast reactors

Lead-bismuth eutectic (LBE) is a candidate coolant for fast reactors. Its physical, chemical and neutronic properties make it a prospect coolant for the reactors. However, corrosion of structure is the main problem of utilizing LBE as a coolant. Compatibility of welded structure with LBE at high temperature should be considered clearly for feasibility of lead-bismuth-cooled fast reactors. This study was preformed to investigate the mechanical properties and corrosion characteristics of the welded ferritic-martensitic (FM) steel, HCM12A, in LBE at 650 °C for 500 h. The welding methods were TIG welding (137 mm/min; 480 W), YAG laser welding (780 mm/min; 287 W) and electron beam welding (1000 mm/min; 60 kW). The oxygen concentration of test environment was maintained at 7 x 10⁻⁷ wt% by injecting Ar-H2-steam gas mixture. Vickers hardness test and SEM/EDX analysis were conducted on the cross section of welded HCM12A. It was found that oxide layer was larger in the weld zones than base metal (BM). However, outer layer was detached on some areas.     

The user model-based summarize and refine approach improves information presentation in spoken dialog systems

A common task for spoken dialog systems (SDS) is to help users select a suitable option (e.g., flight, hotel, and restaurant) from the set of options available. As the number of options increases, the system must have strategies for generating summaries that enable the user to browse the option space efficiently and successfully. In the user-model based summarize and refine approach (UMSR, Demberg and Moore, 2006), options are clustered to maximize utility with respect to a user model, and linguistic devices such as discourse cues and adverbials are used to highlight the trade-offs among the presented items. In a Wizard-of-Oz experiment, we show that the UMSR approach leads to improvements in task success, efficiency, and user satisfaction compared to an approach that clusters the available options to maximize coverage of the domain (Polifroni et al., 2003). In both a laboratory experiment and a web-based experimental paradigm employing the Amazon Mechanical Turk platform, we show that the discourse cues in UMSR summaries help users compare different options and choose between options, even though they do not improve verbatim recall. This effect was observed for both written and spoken stimuli.     

Piezo-Spectroscopic Coefficients of Tetragonal-Prime Yttria-Stabilized Zirconia

The Raman piezo-spectroscopic constants of tetragonal prime yttria-stabilized zirconia under uniaxial compression have been determined. Although the Raman band at 461 cm⁻¹ is not the most intense, it is the most appropriate peak for stress measurement because of its large stress-induced peak shift and the minimum overlap with the other Raman peaks. The peak separation between bands at 258 and 461 cm⁻¹ or 258 and 638 cm⁻¹ can also be used to measure stress.     

Comparison of bioactive components in GABA tea and green tea produced in Taiwan

The aim of this study is to investigate the bioactive components of GABA (γ-aminobutyric acid) tea as compared with green tea produced in Taiwan. Using in total 56 tea samples (28 green tea and 28 GABA tea), moisture content, Hunter L, a and b values, phenolic compounds, amino acids including GABA, fatty acids and ascorbic acid were determined. The results showed that moisture, total free amino acids, crude fat, Hunter L value, total nitrogen, free fatty acids and reducing sugar did not differ significantly between GABA tea and green tea. However, GABA tea had higher Hunter a and b values, while green tea had higher total catechin and ascorbic acid contents (p < 0.05). Of major catechins, epicatechin and epigallocatechin gallate were found to be lower in GABA tea than in green tea. For free amino acids, GABA, alanine, ammonia, lysine, leucine and isoleucine were found to be significantly higher in GABA tea, while the glutamic acid, aspartic acid, and phenylalanine were higher in green tea (p < 0.05). Theanine, tryptophan, valine, threonine and methionine were not found to be different between the two kinds of tea.