We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

We will review some of the major results in random graphs and some of the more challenging open problems. We will cover algorithmic and structural questions. We will touch on newer models, including those related to the WWW.

Random graphs

There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at this https URL

YOLOv4: Optimal Speed and Accuracy of Object Detection

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

The vehicle supply chain industry is in a stage of rapid development with the popularization of electric vehicles. However, counterfeits, also known as fake products, present a non-negligible effect on vehicle supply chains. Counterfeit vehicular parts have created significant losses in both monetary value and social terms. For example, tainted and sub-standard tires or batteries used in vehicles even have caused traffic accidents or loss of human life. Due to these dominant reasons, many researchers have focused on building anti-counterfeiting supply chain systems. More recently, several blockchain-based solutions have been presented, whereas most of them lack privacy protection. In this paper, aiming at the issue, we propose a reliable and Anti-counterfeiting and Privacy-Preserving Blockchain-based vehicle supply chains, called APPB. APPB is characterized by offering dual advantages, by utilizing the immutability of blockchain and keeping the privacy of product selling information and business relationship. Both security analysis and experimental results are conducted, and the results indicate that APPB can achieve privacy protection and anti-counterfeiting with acceptable efficiency.

APPB: Anti-Counterfeiting and Privacy-Preserving Blockchain-Based Vehicle Supply Chains

Study on the mechanical strength and iodine adsorption behavior of coal-based activated carbon based on orthogonal experiments

Dalian Island is located in the sea area near Pingtan County, Fujian, Southeast China. The sea area used to be the junction of the eastern and western ship routes on the Maritime Silk Road, and is also an important region for underwater archaeology in China. This study focused on a sauce-glazed ware of the Song Dynasty, with serious degradation, which was salvaged out of the water at the Dalian Island Wharf. Optical microscopy, scanning electron microscopy, X-ray diffraction analysis, and micro-Raman spectroscopy were used to comprehensively analyze the composition, phase attributes and microstructure of the ware and the surface-attached coagula. The findings revealed that the sea wave-borne debris scoured the surface of the ware, causing mechanical damage to varying degrees and a significant decrease in its degradation resistance. This was the primary factor accounting for the poor preservation state of the salvaged ceramic ware, and the precondition for the subsequent attachment of marine organisms and the deposition of inorganic pollutants. The calcareous skeletons formed on the surface induced by the bio-mineralization of coralline algae (a type of marine plant) could resist the mechanical action caused by the motion of sea waves, thereby slowing down the ware’s degradation process. In other words, the calcareous skeletons played a ‘bio-protective’ role to a certain degree. In addition, inorganic pollutants represented by iron rusts also participated in the corrosion of the glaze. Some pollutants were directly deposited on the pits and cracks on the surface of the ware, which brought stress to the glaze and glaze/body interface, causing the glaze to further crack and spall. Moreover, iron rusts reacted with the glaze, leading to chemical alteration, accompanied by the formation of iron silicate as the alteration product. Anorthite crystals in the interlayer did not participate in the reaction but remained at the original position. The alteration product gradually replaced the original glass phase of the glaze and entered into the body via pores and cracks. In conclusion, the complex degradation morphology of the salvaged sauce-glazed ware could be attributed to the combined action of mechanical damage, marine bio-fouling, and chemical alteration.

/pdf/degradation-mechanism-of-a-sauce-glazed-ware-of-the-song-1hhijpcq.pdf

Degradation Mechanism of a Sauce-Glazed Ware of the Song Dynasty Salvaged out of the Water at Dalian Island Wharf: Part I—The Effect of the Surface-Attached Composite Coagula

Two-dimensional (2D) Nb-based oxynitrides are promising visible-light-responsive photocatalysts for the water splitting reaction, but their photocatalytic activity is degraded by the formation of reduced Nb5+ species and O2− vacancies. To understand the influence of nitridation on the formation of crystal defects, this study synthesized a series of Nb-based oxynitrides through the nitridation of LaKNaNb1−xTaxO5 (x = 0, 0.2, 0.4, 0.6, 0.8, 1.0). During nitridation, K and Na species volatilized, which helped transform the exterior of LaKNaNb1−xTaxO5 into a lattice-matched oxynitride shell. Ta inhibited defect formation, yielding Nb-based oxynitrides with a tunable bandgap between 1.77 and 2.12 eV, straddling the H2 and O2 evolution potentials. After loading with Rh and CoOx cocatalysts, these oxynitrides exhibited good photocatalytic activity for H2 and O2 evolution in visible light (650–750 nm). The nitrided LaKNaTaO5 and LaKNaNb0.8Ta0.2O5 delivered the maximum H2 (19.37 μmol h−1) and O2 (22.81 μmol h−1) evolution rates, respectively. This work provides a strategy for preparing oxynitrides with low defect densities and demonstrates the promising performance of Nb-based oxynitrides for water splitting.

/pdf/defect-engineering-of-two-dimensional-nb-based-oxynitrides-2gaax34s.pdf

Defect engineering of two-dimensional Nb-based oxynitrides for visible-light-driven water splitting to produce H2 and O2

Background: GPC5 rs2352028 is associated with the risk of lung cancer, but its relationship with lung cancer prognosis is unclear. Materials & methods: The authors collected blood samples from 888 patients with lung cancer and used a Cox proportional hazards model to analyze the association between prognosis and GPC5 polymorphism rs2352028. Results: GPC5 rs2352028 C > T was associated with a better prognosis. Patients with CT genotype had longer overall survival than those with CC genotype. Additionally, older and early-stage patients with CT + TT genotype had a lower risk of death than those with CC genotype. Conclusion: GPC5 rs2352028 C > T may play a protective role in patients with lung cancer and GPC5 rs2352028 may be a potential genetic marker for lung cancer prognosis.

Chang Xu

Papers

APPB: Anti-Counterfeiting and Privacy-Preserving Blockchain-Based Vehicle Supply Chains

Study on the mechanical strength and iodine adsorption behavior of coal-based activated carbon based on orthogonal experiments

Degradation Mechanism of a Sauce-Glazed Ware of the Song Dynasty Salvaged out of the Water at Dalian Island Wharf: Part I—The Effect of the Surface-Attached Composite Coagula

Defect engineering of two-dimensional Nb-based oxynitrides for visible-light-driven water splitting to produce H2 and O2

Prognostic value of GPC5 polymorphism rs2352028 and clinical characteristics in Chinese lung cancer patients.