Search

Breast cancer is one of the primary causes of cancer death in the world and has a great impact on women’s health. Generally, the majority of classification methods rely on the high-level feature. However, different levels of features may not be positively correlated for the final results of classification. Inspired by the recent widespread use of deep learning, this study proposes a novel method for classifying benign cancer and malignant breast cancer based on deep features. First, we design Sliding + Random and Sliding + Class Balance Random window slicing strategies for data preprocessing. The two strategies enhance the generalization of model and improve classification performance on minority classes. Second, feature extraction is based on the AlexNet model. We also discuss the influence of intermediate- and high-level features on classification results. Third, different levels of features are input into different machine-learning models for classification, and then, the best combination is chosen. The experimental results show that the data preprocessing of the Sliding + Class Balance Random window slicing strategy produces decent effectiveness on the BreaKHis dataset. The classification accuracy ranges from 83.57% to 88.69% at different magnifications. On this basis, combining intermediate- and high-level features with SVM has the best classification effect. The classification accuracy ranges from 85.30% to 88.76% at different magnifications. Compared with the latest results of F. A. Spanhol’s team who provide BreaKHis data, the presented method shows better classification performance on image-level accuracy. We believe that the proposed method has promising good practical value and research significance.

Light-driven proton directional transport is important in living beings as it could subtly realize the light energy conversion for living uses. In the past years, 2D materials-based nanochannels have shown great potential in active ion transport due to controllable properties, including surface charge distribution, wettability, functionalization, electric structure, and external stimuli responsibility, etc. However, to fuse the inorganic materials into bio-membranes still faces several challenges. Here, we proposed peptide-modified WS

Nanofluidic ion transport through van der Waals heterostructures, composed of two or more types of reconstructed 2D nanomaterials, gives rise to fascinating opportunities for light-energy harvesting, due to coupling between the optoelectronic properties of the layered constituents and ion transport in between the atomic layers. Here, a photoinduced active ion transport phenomenon through transition metal dichalcogenides (TMDs)-based van-der-Waals-like multilayer heterostructures is reported for harnessing ionic power from equilibrium electrolyte solution. The binary heterostructure comprises sequentially stacked 2D-WS2 and 2D-MoS2 multilayers with sub-1 nm interlayer spacing. Upon visible-light illumination, a net ionic flow is initiated through the Janus membrane, suggesting a directional cationic transport from WS2 to MoS2 part. The transport mechanism is explained in terms of a photovoltaic effect due to type II band alignment of WS2 /MoS2 heterostructures. The driving mechanism can be generally applied to a variety of heterogeneous TMD membranes with type II semiconductor heterojunctions. In equilibrium ionic solutions, the maximum ionic photoresponse approaches ≈21 µA cm-2 and ≈45 mV under one sun equivalent excitation. Under optimized conditions, the harvested power density reaches 2 mW m-2 . The proof-of-concept demonstration of photonic-to-ionic power generation within angstrom-scale confinement anticipates potential for light-controlled ionic circuits, artificial photosynthesis, and biomimetic energy conversion.

The emergence of Internet connectivity has led to a significant increase in the volume and complexity of cyber attacks. Abnormal behavior detection systems are valuable tools for ensuring the security in computer networks. However, due to the huge amount and ever increasing diversity of the intrusions, the existing intrusion detection systems, which use machine learning techniques to learn a classifier based on a handcrafted feature vector, are not robust enough to detect sophisticated attacks which cause a high false alarm rate. Therefore, building a flexible in-depth defense system to detect abnormal behavior requires an ability to automatically learn powerful features and analyze large amounts of network traffic. To address these concerns, this paper proposes a novel distributed approach for the detection of abnormal behavior in large-scale networks. The developed model discovers the abnormal behavior from large-scale network traffic data using a combination of a deep feature extraction and multi-layer ensemble support vector machines (SVMs) in a distributed way. First, we perform a non-linear dimensionality reduction, achieved through a distributed deep belief networks on large-scale network traffic data. Then, the obtained features are fed to the multi-layer ensemble SVM. The construction of the ensemble is accomplished through the iterative reduce paradigm based on Spark. Empirical results show a promising gain in performance compared with other existing models.

The bottom-up assembly of ion-channel-mimetic nanofluidic devices and materials with two-dimensional (2D) nano-building blocks paves a straightforward way towards the real-world applications of the novel transport phenomena on a nano- or sub-nanoscale. One immediate challenge is to provide the 2D nanofluidic systems with adaptive responsibilities and asymmetric ion transport characteristics. Herein, we introduce a facile and general strategy to provide a graphene-oxide-based 2D nanofluidic system with photo-switchable ionic current rectification (ICR). The degree of ICR can be prominently enhanced upon UV irradiation and it can be perfectly retrieved under irradiation with visible light. A maximum ICR ratio of about 48 was achieved. The smart and functional nanofluidic devices have applications in energy conversion, chemical sensing, water treatment, etc.

This paper investigates a new attack method to gait recognition systems. Different from typical spoofing attacks that require impostors to mimic certain clothing or walking styles, it proposes to intercept the video stream captured by the on-site camera and replace it with synthesized samples. To this end, we present a novel Generative Adversarial Network (GAN) based approach, which is able to render a faked video from the source walking sequence of a specified subject and the target scene image with both good visual effects and sufficient discriminative details. A new generator architecture is built, where the features of the source foreground sequence and the target background image are combined at multiple scales, making the synthesized video vivid. To fool recognition systems, the silhouette-conditioned losses are specially designed to constrain the static and dynamic consistency between the subjects in the source and generated videos. The person re-identification similarity based triplet loss is exploited to guide the generator, which keeps the personalized appearance properties stable. The edge and flow-related losses further regulate the generation of the attacking video. Two state-of-the-art gait recognition systems are used for evaluation, namely GaitSet and CNN-Gait, and we analyze their performance under attacking. Both the visual fidelity and attacking ability of the generated videos validate the effectiveness of the proposed method.

According to the layout characteristics of parking in the parking lot, the problem of converting the path choice of parking is raised to seek the optimal path and generate the optimal parking guiding line. This paper proposed the concept of node busy factor, and a mathematical model of the optimal path with constraints. The state transition rules of basic ant colony algorithm is improved as well as updating rules of pheromone. After that, the taboo strategy is introduced for further optimization. At last, the experimental results and Matlab simulation verify the effectiveness of the Algorithm.

The scheme design of data collection for Underwater Acoustic Sensor Networks (UASNs) poses many challenges due to long propagation, high mobility, limited bandwidth, multi-path and Doppler Effect. In this paper, unlike the traditional underwater sensor network architecture (single sink or multi-sink), we proposed a novel underwater sensor cloud system based on fog computing in view of time-critical underwater applications. In such an architecture, fog nodes with great computation and storage capacity are responsible for computing, dimension reduction and redundant removal for data collected from physical sensor nodes, and then transfer the processed and compressed data to surface center sink node. After that, the center sink sends the received data from fog nodes to cloud computing center. In addition, in this paper we present distance difference and waiting area-based routing protocol, called DDWA. Finally, in comparison with RDBF, naive flooding and HH-VBF, we conduct extensive simulations using NS-3 simulator to verify the effectiveness and validity of the proposed data collection scheme in the context of the proposed architecture.

In opportunistic networks, selfish nodes will refuse to forward messages for others to save their precious resources such as bandwidth, buffer and energy, and the system performance will degrade significantly. Therefore, it is necessary to motivate the selfish nodes to cooperatively forward messages. However, current cooperation mechanisms for opportunistic networks mainly focus on encouraging nodes to participate in message forwarding, but fail to consider the node congestion problem. When many messages are forwarded to the nodes with high connection degree, these nodes will become congested and discard most messages, which will seriously degrade the routing performance. To stimulate the selfish nodes with limited buffer to cooperatively forward messages, this paper proposes a congestion-aware node cooperation mechanism based on double auction, called CANCMDA. In CANCMDA, nodes first determine the self congestion degree according to current free buffer and message receiving speed. Then, combining message delivery probability and congestion degree, nodes trade messages based on double auction model. The double auction trade process is a bayesian game, and nodes get the optimal bid by solving the bayesian equilibrium to trade messages. The experimental results show that CANCMDA can effectively stimulate selfish nodes to cooperatively forward messages when congested, and achieves higher message delivery ratio with lower overhead ratio, compared with other mechanisms.

Mangroves are an important terrestrial carbon reservoir with numerous ecosystem services. Yet, it is difficult to inventory mangroves because of their low accessibility. A sampling approach that produces accurate assessment while maximizing logistical integrity of inventory operation is often required. Spatial decision support systems SDSSs provide support for integrating such a sampling design of fieldwork with operational considerations and evaluation of alternative scenarios. However, this fieldwork design driven by SDSS is often computationally intensive and repetitive. In this study, we develop a cyber-enabled SDSS framework to facilitate the computationally challenging fieldwork design that requires the efficacious selection of base camps and plots for the inventory of mangroves. Our study area is the Zambezi River Delta, Mozambique. Cyber-enabled capabilities, including scientific workflows and cloud computing, are integrated with the SDSS. Scientific workflows enable the automation of data and modeling tasks in the SDSS. Cloud computing offers on-demand computational support for interoperation among stakeholders for collaborative scenario evaluation for the fieldwork design of mangrove inventory. Further, this framework allows for harnessing high-performance computing capabilities for accelerating the fieldwork design. The cyber-enabled framework provides significant merits in terms of effective coordination among science and logistical teams, assurance of meeting inventory objectives, and an objective basis to collectively and efficaciously evaluate alternative scenarios.

Mangrove forests are highly productive and have large carbon sinks while also providing numerous goods and ecosystem services. However, effective management and conservation of the mangrove forests are often dependent on spatially explicit assessments of the resource. Given the remote and highly dispersed nature of mangroves, estimation of biomass and carbon in mangroves through routine field-based inventories represents a challenging task which is impractical for large-scale planning and assessment. Alternative approaches based on geospatial technologies are needed to support this estimation in large areas. However, spatial data processing and analysis approaches used in this estimation of mangrove biomass and carbon have not been adequately investigated. In this study, we present a spatially explicit analytical framework that integrate remotely sensed data and spatial analyses approaches to support the estimation of mangrove biomass and carbon stock and their spatial patterns in West Africa. Forest canopy height derived from SRTM and ICESat/GLAS data was used to estimate mangrove biomass and carbon in nine West African countries. We developed a geospatial software toolkit that implemented the proposed framework. The spatial analysis framework and software toolkit provide solid support for the estimation and relative comparisons of mangrove-related metrics. While the mean canopy height of mangroves in our study area is 10.2 m, the total biomass and carbon were estimated as 272.56 and 136.28 Tg. Nigeria has the highest total mangrove biomass and carbon in the nine countries, but Cameroon is the country with the largest mean biomass and carbon density. The resulting spatially explicit distributions of mangrove biomass and carbon hold great potential in guiding the strategic planning of large-scale field-based assessment of mangrove forests. This study demonstrates the utility of online geospatial data and spatial analysis as a feasible solution for estimating the distribution of mangrove biomass and carbon at larger or smaller scales.

Nacre-mimetic 2D nanofluidic materials with densely packed sub-nanometer-height lamellar channels find widespread applications in water-, energy-, and environment-related aspects by virtue of their scalable fabrication methods and exceptional transport properties. Recently, light-powered nanofluidic ion transport in synthetic materials gained considerable attention for its remote, noninvasive, and active control of the membrane transport property using the energy of light. Toward practical application, a critical challenge is to overcome the dependence on inhomogeneous or site-specific light illumination. Here, asymmetric photonic-ionic devices based on kirigami-tailored graphene oxide paper are fabricated, and directional nanofluidic ion transport properties therein powered by full-area light illumination are demonstrated. The in-plane asymmetry of the graphene oxide paper is essential to the generation of photoelectric driving force under homogeneous illumination. This light-powered ion transport phenomenon is explained based on a modified carrier diffusion model. In asymmetric nanofluidic structures, enhanced recombination of photoexcited charge carriers at the membrane boundary breaks the electric potential balance in the horizontal direction, and thus drives the ion transport in that direction under symmetric illumination. The kirigami-based strategy provides a facile and scalable way to fabricate paper-like photonic-ionic devices with arbitrary shapes, working as fundamental elements for large-scale light-harvesting nanofluidic circuits.

Very high resolution digital elevation models (DEM) provide the opportunity to represent the micro-level detail of topographic surfaces, thus increasing the accuracy of the applications that are depending on the topographic data. The analyses of micro-level topographic surfaces are particularly important for a series of geospatially related engineering applications. However, the generation of very high resolution DEM using, for example, LiDAR data is often extremely computationally demanding because of the large volume of data involved. Thus, we use a high-performance and parallel computing approach to resolve this big data-related computational challenge facing the generation of very high resolution DEMs from LiDAR data. This parallel computing approach allows us to generate a fine-resolution DEM from LiDAR data efficiently. We applied this parallel computing approach to derive the DEM in our study area, a bottomland hardwood wetland located in the USDA Forest Service Santee Experimental Forest. Our study demonstrated the feasibility and acceleration performance of the parallel interpolation approach for tackling the big data challenge associated with the generation of very high resolution DEM.

In the past decade, the emergence of cyberinfrastructure has rendered high-performance computing resources and parallel technologies increasingly open to domain-specific science discovery. The capability of these high-performance computing resources and associated parallel technologies has greatly stimulated researchers to utilize them for domain-specific problem-solving that requires considerable computational support. The objective of this paper is to investigate in detail the utility of parallel computing in geocomputational modeling. We discuss fundamentals in parallel computing and relevant technologies. To best leverage diverse high-performance computing resources often requires well-crafted parallel computing strategies or algorithms. We review the use of parallel computing for geocomputational modeling by focusing on four aspects: spatial statistics, spatial optimization, spatial simulation, and cartography and geovisualization. We design a case study of a spatial agent-based model to show how parallel computing can be exploited to empower advanced geocomputational modeling. Results demonstrate that the evolving parallel computing provides solid support for computationally intensive geocomputational modeling.

Most of the intrusion detection models (IDM) are constructed with off-line training data. Time-variance characteristic of the practical network system cannot be embodied in the off-line constructed IDM. On-line updating of the off-line IDM with the valued new samples is very necessary. In this paper, a new on-line instruction detection model based on approximate linear dependent (ALD) condition with linear latent feature extraction is proposed to address this problem. Specifically, the valued samples which can represent drift of the practical network are indentified with ALD and prior knowledge. Then, these selected samples are used to update the off-line IDM based on on-line latent feature extraction method and fast machine learning algorithm with sample-based updating strategy. Experiments based on KDD99 data are used to validate the proposed approach.

Sensitivity analysis is an important step in agent-based modeling of complex adaptive spatial systems to evaluate the contribution of influential variables to model response. Sensitivity analysis of agent-based models is computationally demanding, however, and this analysis tends to be intractable for large agent-based modeling. This computational challenge greatly limits our ability to investigate complex spatial dynamics using large agent-based models. The objective of this study is to gain insight into this computational issue by focusing on the sensitivity analysis of large agent-based modeling of spatial opinion exchange, accelerated using multiple graphics processing units (GPUs). We present a heterogeneous parallel computing approach based on nested parallelism for the global sensitivity analysis of the model. The agent-based opinion model is parallelized using many-core GPUs for the simulation of a large number of spatially aware and interacting agents. These agents exchange opinions for developin...

Biological electrogenic systems use protein-based ionic pumps to move salt ions uphill across a cell membrane to accumulate an ion concentration gradient from the equilibrium physiological environment. Toward high-performance and robust artificial electric organs, attaining an antigradient ion transport mode by fully abiotic materials remains a great challenge. Herein, a light-driven proton pump transport phenomenon through a Janus graphene oxide membrane (JGOM) is reported. The JGOM is fabricated by sequential deposition of graphene oxide (GO) nanosheets modified with photobase (BOH) and photoacid (HA) molecules. Upon ultraviolet light illumination, the generation of a net protonic photocurrent through the JGOM, from the HA-GO to the BOH-GO side, is observed. The directional proton flow can thus establish a transmembrane proton concentration gradient of up to 0.8 pH units mm-2 membrane area at a proton transport rate of 3.0 mol h-1 m-2 . Against a concentration gradient, antigradient proton transport can be achieved. The working principle is explained in terms of asymmetric surface charge polarization on HA-GO and BOH-GO multilayers triggered by photoisomerization reactions, and the consequent intramembrane proton concentration gradient. The implementation of membrane-scale light-harvesting 2D nanofluidic system that mimics the charge process of the bioelectric organs makes a straightforward step toward artificial electrogenic and photosynthetic applications.

With the rapid proliferation of wireless networks and mobile devices, Mobile Peer-to-Peer (MP2P) networks have attracted a huge amount of users for resource sharing. However, in an MP2P network, peers frequently join and leave the network, which makes the network topology dynamically change. Thus, it is difficult to establish long term and effective trust relationship among peers, making trust management become a challenging task. In this paper, we propose a dynamic grouping-based trust model DGTM, to classify peers. A group is formed according to the peers' interests. The experiments illustrate that our proposed dynamic grouping-based trust model DGTM always achieves the highest successful transaction rate under different circumstances.

Although many efforts have been made to study the reputation evaluation of node performance in Mobile Peer-to-Peer Networks (MP2P), the existing methodology of node reputation evaluation are not applicable to the scenarios in dynamic networks. To improve the performance of MP2P, in this study, we develop a new evaluation of node reputation by combining an analytic hierarchy process with a fuzzy comprehensive evaluation technology. The reputation of nodes is calculated using a membership function to account for direct and indirect trust, residual energy and fuzzy comprehensive evaluation. The experimental results show that the proposed method is not only robust in delivering a ratio of packet, but also reduce the average latency and overload ratio in MP2P environment.

Previous studies suggested that the development of producer service resulted from manufacturing outsourcing. On the other hand, competitive producer service can promote the progress of manufacturing industry. Theoretically, correlation may exist between them. According to co-integration theory, this paper empirically tests the time series data col-lected from both producer service and manufacturing industry in Shanghai, indicating that co-integration does exist between producer service and local manufacturing industry. Empirical results of this paper partially support the conclusions of prior studies addressed above.