Search

Waste particleboards are largely generated from discarded furniture, floor, and other building materials, which can to be efficiently utilized as energy source by pyrolysis technologies. However, due to the existence of the adhesives and ash inside, waste particleboards are relatively hard to be pyrolyzed and form certain amounts of nitrogenous compounds. In order to improve the pyrolysis reactivity and tune the pyrolytic products, wood is proposed to be blended with waste particleboards for the co-pyrolysis treatment. In this study, the pyrolysis characteristics and chemical composition of pyrolytic vapors of samples with different blending ratios were studied using the TGA and Py-GC-MS analyses. The results showed as increasing the weight percentage of poplar wood, the pyrolysis activation energy and char residue content of the blends were gradually reduced, which was almost a linear correlation. Nevertheless, the chemical composition and distribution of pyrolytic vapors products of the samples was varied nonlinearly with blending ratio, and poplar wood exhibited a significant influence on the production of pyrolytic compounds from waste particleboards. The yields of most pyrolysis products reached the highest when the ratio of poplar wood was 50%, which probably indicated the existence of a synergy effect during the co-pyrolysis of different types of feedstock. It is believed that the results of this study are beneficial to the pyrolysis conversion of waste wood-based materials.

Recommender system has recently attracted a lot of attention in the information service community. Currently, most recommendation models use deep neural networks to learn user preferences for items and make the final recommendations. However, these current models have not effectively captured the deep semantic features of users and items and have not fully used the auxiliary information of the items. This may result in unsatisfactory recommendations for users’ projects. In order to solve the above problem, in this paper, a novel recommendation model called RM-DRL (Recommendation Model based on Deep Representation Learning) was proposed. It mainly consists of two modules: Information Preprocessing and Feature Representation. The former generates the user’s primitive feature vectors and the items used in the latter. The latter consists of two phases: Representation Learning for Item Features (RL-IF) and Representation Learning for User Features (RL-UF). The RL-IF takes the primitive feature vectors of the item as input and uses a multi-layer Convolutional Neural Network (CNN) to learn to accurately produce the semantic feature vector of the item through multi-task learning. In RL-UF, the user primitive feature vectors and semantic feature vectors of the user preference history, and the positive and negative items were taken as input, and a novel Attention-Integrated Gated Recurrent Unit (AIGRU) neural network was proposed to learn to accurately produce user semantic feature vector. After the Feature Representation module converges, the semantic feature vectors of the users and the items can be used to calculate the users’ preferences on the items via vector dot product. Extensive experiments on five real-world datasets show that RM-DRL remarkably outperforms state-of-the-art baselines in solving the recommendation problem.

In recent years, group recommendation has become a research hotspot and focus in online social network community. Currently, several deep-learning-based approaches are leveraged to learn preferences of groups for items and predict the next items in which groups may be interested. Yet, their recommendation performance is still unsatisfactory due to the sparse group–item interactions. In order to address this problem, in this article, we introduce an effective model, namely Social Group Recommendation model with TrAdaBoost (SGRTAB), to raise the performance of group recommendation in online social networks. The SGRTAB model includes two stages: data preprocessing (DP) and model optimization (MO). In DP, SGRTAB produces inputs for MO and implements three related tasks: extracting individual features, handling group data via GloVe, and utilizing user contribute ratings to their own groups, whereas in MO, SGRTAB implements group preference learning with the assistance of user preference learning based on the TrAdaBoost algorithm. Specifically, SGRTAB can effectively absorb the knowledge of user preferences into the process of group preference learning through the idea of transferring-ensemble learning. Moreover, extensive experiments on four real-world data sets indicate that the proposed SGRTAB model significantly outperforms the state-of-the-art baselines for social group recommendation.

Maintaining network connectivity is an important challenge for vehicular ad hoc network (VANET) in an urban scene, which has more complex road conditions than highways and suburban areas. Most existing studies analyze end-to-end connectivity probability under a certain node distribution model, and reveal the relationship among network connectivity, node density, and a communication range. Because of various influencing factors and changing communication states, most of their results are not applicable to VANET in an urban scene. In this article, we propose a connectivity prediction-based dynamic clustering (DC) model for VANET in an urban scene. First, we introduce a connectivity prediction method (CP) according to the features of a vehicle node and relative features among vehicle nodes. Then, we formulate a DC model based on connectivity among vehicle nodes and vehicle node density. Finally, we present a DC model-based routing method to realize stable communications among vehicle nodes. The experimental results show that the proposed CP can achieve a lower error rate than the geographic routing based on predictive locations and multilayer perceptron. The proposed routing method can achieve lower end-to-end latency and higher delivery rate than the greedy perimeter stateless routing and modified distributed and mobility-adaptive clustering-based methods.

Location-based recommendation has recently received a lot of attention in the communities of information service and mobile application. Its task is to provide personalized recommendations of points of interest (POIs) to users at a certain time and location. However, existing location-based recommendation models have at least two main drawbacks: first they cannot adequately capture semantic features of POIs and users, which may lead to unsatisfactory recommendations and second they cannot effectively address the cold-start problem. To address the above drawbacks, in this article, we first propose a novel deep representation learning-based model (DRLM) for improving the recommendation accuracy. In DRLM, we mainly focus on learning to accurately represent semantic features of POIs and users. Specifically, four co-occurrence matrices are constructed to produce four different original features for each POI, and a principal component analysis (PCA) algorithm is utilized to generate a semantic feature of each POI from its four original features. On the other hand, a three-modal simple recurrent unit (TMSRU) network is given to constructed semantic features of users using semantic features of POIs, times, and locations. We further propose minimum description length (MDL)-based and skyline-based strategies to address the cold-start issues for new users and new POIs, respectively. Through experiments on two real-world data sets, we show that compared with the state-of-the-art approaches, the proposed model DRLM can achieve the superior performance in terms of high recommendation accuracy and effectiveness in handling the cold-start problem.

Collective action propagation, which can be as large as billions of people adopting Facebook or as small as a few researchers citing a paper, exists in various real-life scenarios. Here, we perform a large-scale investigation of collective action propagation with “recurrence” phenomena. We consider actions that propagate in a social network with multiple communities and find the growth in the propagation breadth of collective action can be explained by a simple mathematical model with an analytical solution. We use datasets on the growth of total views of TED and YouTube videos, the prize pool of Dota 2 tournaments, and a total gross of movies to investigate collective action propagation with recurrence phenomena. Experimental results reveal that our model can capture universal features of collective action propagation, validating the idea that collective action propagation with recurrence results from an action being transmitted from communities to communities.

Driven by the vision of Internet of Things, some research efforts have already focused on designing a network of efficient speech recognition for the development of edge computing. Other researches (such as tpool2) do not make full use of spatial and temporal information in the acoustic features of speech. In this paper, we propose a compact speech recognition network with spatio-temporal features for edge computing, named EdgeRNN. Alternatively, EdgeRNN uses 1-Dimensional Convolutional Neural Network (1-D CNN) to process the overall spatial information of each frequency domain of the acoustic features. A Recurrent Neural Network (RNN) is used to process the temporal information of each frequency domain of the acoustic features. In addition, we propose a simplified attention mechanism to enhance the portion of the network that contributes to the final identification. The overall performance of EdgeRNN has been verified on speech emotion and keywords recognition. The IEMOCAP dataset is used in speech emotion recognition, and the unweighted average recall (UAR) reaches 63.98%. Speech keywords recognition uses Google's Speech Commands Datasets V1 with a weighted average recall (WAR) of 96.82%. Compared with the experimental results of the related efficient networks on Raspberry Pi 3B+, the accuracies of EdgeRNN have been improved on both of speech emotion and keywords recognition.

The recommender system has recently drawn a lot of attention to the communities of information services and mobile applications. Many deep learning-based recommendation models have been proposed to learn the feature representations from items. However, in Internet of Things (IoT), items’ description information are typically heterogeneous and multimodal, posing a challenge to items’ representation learning of recommendation models. To address this challenge and to improve the recommendation effectiveness in IoT, a novel multimodal representation learning-based model (MRLM) has been proposed. In MRLM, two closely related modules were trained simultaneously; they are global feature representation learning and multimodal feature representation learning. The former was designed to learn to accurately represent the global features of items and users through simultaneous training on three tasks: 1) triplet metric learning; 2) softmax classification; and 3) microscopic verification. The latter was proposed to refine items’ global features and to generate the final multimodal features by using items’ multimodal description information. After MRLM converged, items’ multimodal features and users’ global features could be used to calculate users’ preferences on items via cosine similarity. Through extensive experiments on two real-world datasets, MRLM remarkably improved the recommendation effectiveness in IoT.

Using large-scale GPS trajectory data to improve taxi services has recently attracted much attention in Internet of Things and smart city communities. In this paper, we use a large-scale GPS trajectory dataset generated by over 12 000 taxis in a period of three months in Shanghai, China, and present an efficient passenger-hunting recommendation framework with the multitask deep learning paradigm. This framework contains two modules: 1) offline training of passenger-hunting recommendation model (OT-PHRM) and 2) online application of passenger-hunting recommendation model (OA-PHRM). The module OT-PHRM mainly includes two deep convolutional neural networks (DCNNs) and uses the multitask learning strategy. The first DCNN realizes the region prediction for picking up passengers, while the second DCNN uses the weight-sharing structure to predict the levels of road congestion and earnings of carrying passengers. In particular, for the input of two DCNNs, we not only consider contextual features of taxi driving, region features and valuable statistical features, but also combine individual features into meaningful ones. In the module OA-PHRM, we propose DL-PHRec, which calculates three prediction values using two trained DCNNs in OT-PHRM in real time, and then recommends a personal ranking-list of regions to each taxi driver according to their scores. The experimental results show the feasibility and effectiveness of our recommendation framework.

It is an important challenge to detect an overlapping community and its evolving tendency in a complex network. To our best knowledge, there is no such an overlapping community detection method that exhibits high normalized mutual information (NMI) and ${F}$ -score, and can also predict an overlapping community’s future considering node evolution, activeness, and multiscaling. This paper presents a novel method based on node vitality, an extension of node fitness for modeling network evolution constrained by multiscaling and preferential attachment. First, according to a node’s dynamics such as link creation and destruction, we find node vitality by comparing consecutive network snapshots. Then, we combine it with the fitness function to obtain a new objective function. Next, by optimizing the objective function, we expand maximal cliques, reassign overlapping nodes, and find the overlapping community that matches not only the current network but also the future version of the network. Through experiments, we show that its NMI and ${F}$ -score exceed those of the state-of-the-art methods under diverse conditions of overlaps and connection densities. We also validate the effectiveness of node vitality for modeling a node’s evolution. Finally, we show how to detect an overlapping community in a real-world evolving network.

This paper presents a parameterized batch reinforcement learning algorithm for near-optimal longitudinal control of autonomous land vehicles (ALVs). The proposed approach uses an actor-critic architecture, where parameterized feature vectors based on kernels are learned from collected samples for approximating the value functions and policies. One difference between the parameterized batch actor-critic (PBAC) algorithm and previous actor-critic learning approaches is that the critic and actor in PBAC share the same linear features, which has been theoretically proved to be a beneficial property for the convergence of actor-critic learning approaches. In order to obtain better learning efficiency, least-squares-based batch updating rules are designed for the critic and actor, respectively. Based on the PBAC learning algorithm, a data-driven longitudinal control method is presented for ALVs to obtain near-optimal control policies which adaptively tune the fuel/brake control signals to track different speeds. A multiobjective reward function is designed so that both tracking precision and driving smoothness are considered. Extensive experiments were conducted on a real ALV platform while driving on flat, slippery, sloping, and bumpy roads. The experimental results illustrate the superiority of the PBAC-based self-learning controller over conventional longitudinal control methods such as proportional-integral (PI) control and learning-based PI control.

This study presents a three-dimensional numerical simulation of a circular bottom-sitting OWC device for wave energy extraction. The numerical model is based on Reynolds-Averaged Navier-Stokes equations with a modified k-ω turbulence model and a Volume-of-Fluid method to track the air-water surface. The numerical model and setup are validated against a set of wave-flume experimental results. The numerical simulation provides detailed flow-field information, which allows for an analysis of the spatial non-uniformity inside the OWC chamber and the measurement error associated with it. New numerical results show that vortex shedding at the sharp edge of the OWC chamber enhances the spatial non-uniformity inside the OWC chamber through a resonant sloshing mechanism. The enhanced spatial non-uniformity can affect the accuracy of the experimentally-determined quantities such as the extraction efficiency of the OWC chamber and the characteristics of the power take-off device; Suggestions are also provided to improve the accuracy of the experimentally-determined quantities that may be affected by the spatial non-uniformity inside the OWC chamber.

An information dissemination network (i.e., a cascade) with a dynamic graph structure is formed when a novel idea or message spreads from person to person. Predicting the growth of cascades is one of the fundamental problems in social network analysis. Existing deep learning models for cascade prediction are primarily based on recurrent neural networks and representation on random walks or propagation paths. However, these models are not sufficient for learning the deep spatial and temporal features of an entire cascade. Therefore, a new model, called Cascade2vec, is proposed to learn the dynamic graph representation of cascades based on graph recurrent neural networks. To learn more effective graph-level representation of cascades, the current graph neural networks are improved by designing a graph residual block, which shares attention weights between nodes, and by transforming features through perception layers. Furthermore, the proposed graph neural network is integrated into a recurrent neural network to learn the temporal features between graphs. With this method, both the spatial and temporal characteristics of cascades are learned in Cascade2vec. The experimental results show that our method significantly reduces the mean squared logarithmic error and median squared logarithmic error by 16.1% and 12%, respectively, in the cascade prediction at one hour in the Microblog network dataset compared with strong baselines.

Combined with the Internet of Vehicles, some intelligent systems for connected health can make medical vehicles transport medical supplies more safely and timely in response to catastrophic natural disasters or serious accidents. However, in an urban scenario, the crisscrossing of roads and the uneven distribution of vehicles exist, which lead to problems such as the high mobility of vehicles and the attachment of data. These have become important contributors to the low stability of the vehicle community and the high distortion of the data among medical vehicles. Focusing on the above problems, this paper proposes a prediction-based multirole classification community clustering method (PMRC) for the vehicular ad hoc network (VANET). The experimental results show that the method can effectively improve the stability of the community in VANET and reduce the probability of data distortion.

Recently, deep learning has gained great popularity in the area of recommender systems. Various combinations of deep learning, collaborative recommendation and content-based recommendation have occurred. However, as one of the three most significant recommendation techniques, hybrid recommendation has little cooperation with deep learning. Besides, most current deep hybrid models only incorporate two simple recommendation methods together in post-fusion, leaving massive space for further exploration of better combinations. In this paper, we apply deep learning to hybrid recommendation, proposing a deep hybrid recommendation model DMFL (Deep Metric Factorization Learning). In DMFL, we combine deep learning with improved machine learning models to learn the interaction between users and items from multiple perspectives. Such deep hybrid learning helps to reflect the user preference more comprehensively and strengthen model’s ability of generalization. We also propose a more accurate method of user feature representation, taking both long-term static characteristics and short-term dynamic interest changes of users into consideration. Furthermore, thorough experiments have been conducted on real-world datasets, which strongly proves the effectiveness and efficiency of the proposed model.

Emotional intelligence plays an important role in human intelligence and is a recent research hotspot. With the rapid development of deep learning techniques in recent years, several neural network-based emotional text generation methods have been investigated. However, the existing emotional text generation approaches often suffer from the problem of requiring large-scale annotated data. Generative adversarial network (GAN) has shown promising results in natural language generation and data enhancement. In order to solve the above problem, this paper proposes a GAN-based cross-domain text sentiment transfer model, which uses annotated data from other domains to assist in the training of emotional text generation network. By combining adversarial reinforcement learning with supervised learning, our model is able to extract patterns of sentiment transformation and apply them in emotional text generation. The experimental results have shown that our approach outperforms the state-of-the-art methods and is able to generate high-quality emotional text while maintaining the consistency of domain information and content semantics.

The fundamental roadblock toward commercial-scale wave power operations is cost. The main objective of this work was to address the cost challenge facing wave energy commercialization through cost-sharing with pile breakwaters to be built for shore protection. This was achieved in this study through a dual-functional wave-power plant for generation of wave-power electricity and protection against coastal erosion for sustainable coastal development. The dual-functional wave-power plant was formed by integrating oscillating-water-column (OWC) devices into a pile breakwater, with each pile being an OWC-pile equipped with a power take-off device. The power extraction efficiency and hydrodynamic characteristics of the dual-functional wave-power plant were measured in a wave flume under various wave conditions. An orifice was used at the top of the pneumatic chamber of each OWC-pile to simulate the power take-off device. To evaluate the performance of the power plant in wave power extraction and shore protection, the surface elevation and pressure inside the OWC chamber, as well as the scattered waves, were measured. It was found that comparing to a standalone OWC-pile device with an identical design and geometric characteristics, an OWC-pile in the dual-functional wave-power plant could achieve significantly larger power-extraction efficiency. Comparing to a pile breakwater with the same dimensions, the wave transmission and reflection of the dual-functional wave-power plant were both weaker, especially the wave reflection, which is beneficial for structure safety and shore protection. Based on the Froude’s law of similarity and an estimation of the effect of air compressibility at full scale, an evaluation of the performance of a dual-functional wave-power plant at full scale was also provided. The findings of this work promote close collaboration between wave-energy utilization community and the shore-protection community for commercial-scale deployment of wave energy converters and contribute to making wave energy economically competitive.

Effective evaluation on the coordination between socio-economy (V) and carbon emission (W) is essential for sustainable urban planning, and this study introduces an improved coupling coordination degree (CCD) model for conducting such evaluation. The limitation of the traditional CCD model has been examined. The effectiveness of the improved CCD model is demonstrated with its application in 30 case provinces of China by using the data collected for the period 1995–2015. The results show that: (1) The subjectivity embodied in the traditional CCD model in analyzing the coordination between V and W significantly distorts the true coordination between the two systems. (2) The improved CCD model is effective as its application can reflect the reality of the sample cases. (3) The average coordination degrees of the 30 case provinces for the surveyed period show an upward evolution, and the coordination statuses of these provinces have been transforming from low-grade symbiosis to high-grade symbiosis. (4) The spatial distribution of coupling coordination degree in China is characterized by a higher degree of coordination in the Eastern provinces. The results of applying the improved coupling coordination degree model can help decision makers formulate effective sustainable measures to balance socio-economic development with carbon emission reduction.

Group recommendation research has recently received much attention in a recommender system community. Currently, several deep-learning-based methods are used in group recommendation to learn preferences of groups on items and predict the next ones in which groups may be interested. However, their recommendation effectiveness is disappointing. To address this challenge, this article proposes a novel model called a multiattention-based group recommendation model (MAGRM). It well utilizes multiattention-based deep neural network structures to achieve accurate group recommendation. We train its two closely related modules: vector representation for group features and preference learning for groups on items. The former is proposed to learn to accurately represent each group’s deep semantic features. It integrates four aspects of subfeatures: group co-occurrence, group description, and external and internal social features. In particular, we employ multiattention networks to learn to capture internal social features for groups. The latter employs a neural attention mechanism to depict preference interactions between each group and its members and then combines group and item features to accurately learn group preferences on items. Through extensive experiments on two real-world databases, we show that MAGRM remarkably outperforms the state-of-the-art methods in solving a group recommendation problem.

Skyline query processing has recently received a lot of attention in database and data mining communities. However, most existing algorithms consider how to efficiently process skyline queries from base tables. Obviously, when the data size and the number of skyline queries increase, the time cost of skyline queries will increase exponentially, which will seriously influence the query efficiency. Motivated by the above, in this paper, we consider improving the query efficiency via skyline views and propose a cost-based algorithm (abbr. CA) to efficiently select the optimal set of skyline views for storage. The CA algorithm mainly includes two phases: (i) reduce the skyline views selection to the minimum steiner tree problem and obtain the approximate optimal set AOS of skyline views, and (ii) adjust AOS and produce the final optimal set FOS of skyline views based on the simulated annealing. Moreover, in order to improve the extendibility of the CA algorithm, we implement it based on the map/reduce distributed computation model in cloud computing environments. The detailed theoretical analyses and extensive experiments demonstrate that the CA algorithm is both efficient and effective.

Discovering hidden knowledge patterns in trajectory data can help to hunt passengers for taxi drivers. And it is an important issue in the intelligent transportation domain. However, the existing approaches are inaccurate in real applications. Hence in this paper, by using the GPS trajectory big data of taxis, we innovatively present an efficient and effective recommendation system (TRec) for hunting passengers with deep neural structures. This proposed recommendation system is mainly based on the wide & deep model, which is trained wide linear frameworks and deep neural networks together and can simultaneously have the benefits of memorization and generalization to hunt passengers. Meanwhile, in order to improve the accuracy of hunt passengers, our proposed recommendation system uses experienced taxi drivers as learning objects, while considering the prediction of hunting passengers, the prediction of road condition and the evaluation of earnings simultaneously. A performance study using the real GPS trajectory dataset is conducted to evaluate our proposed recommendation system. The experimental evaluation shows that the proposed recommendation system is both efficient and effective. This work strides forward a first step toward building a recommendation system for hunting passengers based on the wide & deep model.

Effectively estimating the permeability of shale-gas reservoirs by traditional models is challenging; however, study in this area is lacking and deficient. We have developed a method for predicting the permeability of shale-gas reservoirs from porosity and rock compositions including mineralogy and organic matter content, which is applicable to laboratory data and downhole measurements. First, two sets including 38 samples from the Longmaxi Formations were tested for porosity, permeability, grain density, total-organic-carbon (TOC) content, mineralogical composition, and low-temperature nitrogen adsorption (LTNA). We used Kozeny’s equation to calculate the specific surface area, which was viewed as the effective specific surface in shale formations through comparing with the specific surface from LTNA. Furthermore, the effective specific surface was found to be positively correlated with clays, pyrite, and TOC contents, whereas it was negatively correlated with quartz, feldspar, and carbonates. Then, an empirical equation between the effective specific surface area and rock compositions was established via a partial least-squares method, which can process the serious multicollinearity of various mineral contents. Combined with Kozeny’s equation, this equation enabled a prediction of permeability from porosity and rock composition. The results indicated that the predicted and measured permeability have a reasonable match. Compared with other models, this method avoids the correlations between porosity and minerals, providing better insight to the influence of minerals and organic matter on permeability. The influences of rock composition on permeability are different, and are caused by the different types and sizes of pores developed within the minerals and organic matter. In addition, the new method was successfully applied to the well-log data from a shale-gas well for permeability predictions.

Complex networks represented as node adjacency matrices constrains the application of machine learning and parallel algorithms. To address this limitation, network embedding (i.e., graph representation) has been intensively studied to learn a fixed-length vector for each node in an embedding space, where the node properties in the original graph are preserved. Existing methods mainly focus on learning embedding vectors to preserve nodes proximity, i.e., nodes next to each other in the graph space should also be closed in the embedding space, but do not enforce algebraic statistical properties to be shared between the embedding space and graph space. In this work, we propose a lightweight model, entitled Network2Vec, to learn network embedding on the base of semantic distance mapping between the graph space and embedding space. The model builds a bridge between the two spaces leveraging the property of group homomorphism. Experiments on different learning tasks, including node classification, link prediction, and community visualization, demonstrate the effectiveness and efficiency of the new embedding method, which improves the state-of-the-art model by 19% in node classification and 7% in link prediction tasks at most. In addition, our method is significantly faster, consuming only a fraction of the time used by some famous methods., 8 pages. 8 figures. Will appear at workshop on the conference ICDM 2020

In social networks, reputation aggregation is an effective approach for recognizing malicious behaviors and individuals. However, organized collusive fraud to obtain a high reputation is the most common and most harmful type of widespread network attack. Therefore, countering the collusion in reputation aggregation systems and further detecting the collusive fraud organizations (CFO) is a significant challenge. In this study, we propose a multi-criteria detection scheme of collusive fraud organization, named MD-CFO, to identify CFO in social networks. This scheme is based on a new universal reputation aggregation method, which includes the calculation of a reputation score and a universal factor. Moreover, five detection criteria and their corresponding factors, i.e., the rating difference fraud factor, rating frequency fraud factor, collaborative behavior fraud factor, suspicious relationship fraud factor, and CFO factor, are used to evaluate the likelihood of being a colluder. Furthermore, we propose three algorithms for detecting CFO and colluders. To prevent false-positive detection results, a correction mechanism called time slice verification (TSV) is used to certify a nodes likelihood of suspicion or fraud in a series of time slices, thereby excluding honest nodes from CFO detection. Finally, empirical simulations are used to test the feasibility and effectiveness of our scheme. We propose a new universal reputation aggregation method.We present a multi-criteria collusive fraud organization detection scheme for reputation aggregation.Five detection criteria are used to evaluate the likelihood of being a colluder.The scheme can detect both executors and organizers in CFO.A mechanism called time slice verification is addressed to prevent false-positive detection.

The connectivity of a large-scale heterogeneous Internet of Vehicles (IoV) is an important challenge for the environment of urban road scenes, which feature crossroads, buildings, and communication devices. The uneven density and wide distribution of vehicles in the city, the building barriers, and the interference of other communication technology will make the connectivity weaker. To the best of our knowledge, there is no formal method to model and analyze connectivity by considering the environment of urban road scenes. This paper presents such a theoretical method to investigate four connectivity properties- i.e., possibility, data forwarding time, link forwarding capability, and packet error rate-and deduce a connectivity model. Using experiments, we prove that the proposed model can ensure that the connectivity is effective and reliable in an urban road scene, and it can be used to accurately evaluate the network connectivity of a highly dynamic IoV.

Mussel-inspired polymers have attracted interest due to their strong adhesion to various substrates; however, the existing synthetic methods are usually complicated and require the utilization of organic solvents. In this work, a novel mussel-inspired polyamide amine-dihydrocaffeic acid (PPCA) polymeric adhesive was designed by grafting a catechol-containing dihydrocaffeic acid (DHCA) monomer to the polyamide amine (PPC) backbone as an economic commercial agent. Due to interactions between catechol groups and various substrate surfaces, PPCA can be used as a general-purpose adhesive for various substrates and exhibited excellent adhesion to low-energy hydrophobic surfaces (e.g., an adhesive strength of 0.35 ± 0.02 MPa for PTFE). No tedious solidification processes were required to reduce the energy and time consumption of the process while exhibiting long-term adhesion. Epichlorohydrin (EPI) was integrated into the polymer system to generate self-crosslinking to improve the cohesive strength, which further improved its adhesive performance. This adhesive also exhibited excellent wet adhesion because the catechol groups induced synergistic electrostatic and hydrogen-bonding interactions between azetidinium groups and phenols. This facile strategy provides a clean and general method to prepare strong adhesives for structural engineering applications under harsh conditions.

This study examines the performance of a numerical method that introduces an artificial sink term to the Reynolds-averaged Navier Stokes equations to simulate the flow through an orifice used as a quadratic Power Takeoff (PTO) for Oscillating Water Columns (OWCs). The method replaces the quadratic PTO by an artificial Forchheimer-flow region (referred to as the artificial Forchheimer-flow method). The performance of this method was evaluated by making comparisons with the existing experimental results for two OWCs and the numerical results obtained by using air-water two-phase flow simulations of the air flow through an orifice (referred to as the orifice-flow method). The surface elevations, velocity fields and pressure fields obtained by the orifice-flow and artificial Forchheimer-flow methods are compared. To use the artificial Forchheimer-flow method, an equation for specifying the Forchheimer coefficient is provided and the sensitivity of the pneumatic efficiency to the Forchheimer coefficient is discussed. It can be concluded that the artificial Forchheimer-flow method can satisfactorily reproduce the measured pneumatic efficiency, pressure field in the air, the velocity field in the water and the cross-sectional average velocity of the air. Compared to the orifice-flow method, the artificial Forchheimer-flow method can speed up the simulation by at least 25 times.

Taxis provide essential transport services in urban areas. In the taxi industry, the income level remains a cause of concern for taxi drivers as well as regulators. Mining underlying factors affecting the income level will not only benefit the newcomers and low-income drivers but also assist in developing effective optimization algorithms for taxi operations. This paper intends to disclose the factors affecting incomes along with their quantitative influence by mining over 167 million GPS records from nearly 8000 taxis in Shanghai. We first identify a marked difference in drivers’ incomes and categorize drivers into three income levels accordingly. We next investigate the overall search-delivery process, thereby defining several factors that may affect the income level. We then develop a generalized multi-level ordered logit (GMOL) model to find the significant factors that influence incomes. Finally, we compute the elasticity for those significant factors and present their contributions, as well as challenge some preconceived ideas regarding how to earn high incomes.

Determination of the total organic carbon (TOC) from well logs is an important step in formation evaluation of shale reservoirs. The ΔlogR model is one of the most commonly used methods in determining the TOC of the source rocks. Wang et al. (2016) made three revisions on ΔlogR model to enhance the prediction accuracy, one of which is to use a changeable slope of log 10 R t versus porosity logs to replace a fixed value. Based on the original ΔlogR model and Wang's revision, we propose an improved model for a better TOC estimation in this paper. Specifically, the approximate linear baseline is replaced by a theoretical one. Under this condition, both the slope and resistivity and porosity log values of the baselined rocks vary from depth to depth. The determination of theoretical baseline is presented, and the varied baseline log values are obtained by a simple mathematical technique. In addition to the shale play of Western Canada Sedimentary Basin, two study areas of Sichuan Basin and Ordos Basin in China, were used to assess the applications of the proposed model and its improvement over the original model. The analysis and discussions presented in this paper indicate that the improved model offers more reliable results in shale plays with large lithological variations. Also, the TOC content of the three study areas was estimated using the proposed model. The results showed that the TOC values estimated from well logs and core analysis are in good agreement.

Oscillating water column (OWC) devices for wave power extraction are appealing, but are still in need of research. In this study, a series of wave-flume experiments was conducted to examine the hydrodynamic performance of a rectangular OWC device fixed in regular waves. Two types of orifices, slot orifices and circular orifices, were used to simulate the nonlinear power take-off (PTO) mechanism, and the effects of orifice geometry were examined. A two-point measurement method was proposed to reconstruct the instantaneous spatial profile of the water surface inside the OWC chamber for reducing bias in the measured wave power extraction efficiency. The flow characteristics of PTO were described by a quadratic loss coefficient, and our experimental results showed that the quadratic loss coefficient of the slot orifices varied with wave period and slot geometry. Empirical formulas were proposed for the quadratic loss coefficients of the two types of orifices. The ability to determine the quadratic loss coefficient of an orifice will allow us to design orifices for small-scale tests and calculate the power extraction using only pressure measurement. Our results also suggested that the pressure coefficient should be more reliable than the amplification coefficient as an indicator of the power extraction performance of an OWC device.

Feature representation is critical not only for pattern recognition tasks but also for reinforcement learning (RL) methods to solve learning control problems under uncertainties. In this paper, a manifold-based RL approach using the principle of locally linear reconstruction (LLR) is proposed for Markov decision processes with large or continuous state spaces. In the proposed approach, an LLR-based feature learning scheme is developed for value function approximation in RL, where a set of smooth feature vectors is generated by preserving the local approximation properties of neighboring points in the original state space. By using the proposed feature learning scheme, an LLR-based approximate policy iteration (API) algorithm is designed for learning control problems with large or continuous state spaces. The relationship between the value approximation error of a new data point and the estimated values of its nearest neighbors is analyzed. In order to compare different feature representation and learning approaches for RL, a comprehensive simulation and experimental study was conducted on three benchmark learning control problems. It is illustrated that under a wide range of parameter settings, the LLR-based API algorithm can obtain better learning control performance than the previous API methods with different feature representation schemes.

Exponential growth of information generated by social networks requires efficient and scalable recommendation techniques to produce useful results. Traditional methods have become unqualified because they consider only ratings instead of rankings in an item list, and they ignore social contextual information, which is valuable for predicting users’ preference. It is significant and challenging to fuse social contextual information into learning to recommendation methods. In this study, the authors first extend user latent features by exploiting users’ social relationship such as friendship or trust relations, and extend item latent features with concurrent items. Then they integrate both users’ and items’ social contextual information into a pairwise learning to recommendation model (named as UIContextRank) to enhance ranking accuracy and recommendation quality. Furthermore, they extend UIContextRank in a distributed environment to improve efficiency and scalability. The authors conduct experiments on both bidirectional and unidirectional social network datasets. The results show that their method significantly outperforms other approaches.

In general, the stock prices of the same industry have a similar trend, but those of different industries do not. When investing in stocks of different industries, one should select the optimal model from lots of trading models for each industry because any model may not be suitable for capturing the stock trends of all industries. However, the study has not been carried out at present. In this paper, firstly we select 424 S&P 500 index component stocks (SPICS) and 185 CSI 300 index component stocks (CSICS) as the research objects from 2010 to 2017, divide them into 9 industries such as finance and energy respectively. Secondly, we apply 12 widely used machine learning algorithms to generate stock trading signals in different industries and execute the back-testing based on the trading signals. Thirdly, we use a non-parametric statistical test to evaluate whether there are significant differences among the trading performance evaluation indicators (PEI) of different models in the same industry. Finally, we propose a series of rules to select the optimal models for stock investment of every industry. The analytical results on SPICS and CSICS show that we can find the optimal trading models for each industry based on the statistical tests and the rules. Most importantly, the PEI of the best algorithms can be significantly better than that of the benchmark index and "Buy and Hold" strategy. Therefore, the algorithms can be used for making profits from industry stock trading.

MEPCMs have drawn tremendous attention in TES fields. In this study, eutectic of CA and PA was encapsulated by PVC, and the novel MEPCM for TES applications was developed. Microcapsules were prepared by solvent evaporation method and examined using DSC, SEM, LPSA, FTIR, thermal cycling test and TGA. DSC results demonstrated that the phase change temperature of CA-PA eutectic decreased after microencapsulation. The most satisfied sample was the MEPCM with a shell-core ratio of 1:2, which melted at 17.1 °C with the latent heat of 92.1 J/g, the encapsulation ratio was 57.7%. SEM images and LPSA results showed that the prepared MEPCMs were consisted with micro-sized spheres. FTIR results confirmed that the PCM core was successfully encapsulated by the PVC shell, and no chemical reaction was found among the components. The prepared MEPCM showed an excellent thermal reliability after 500 times of thermal cycling. Microencapsulation enhanced the starting temperature of CA-PA eutectic thermal decomposition, and the novel MEPCM exhibited an excellent thermal stability at working temperature from TGA results. Consequently, MEPCM (1:2) was the optimal composition and had great potential for TES applications.

With the booming exploration and development of unconventional hydrocarbon resources in source rocks, the estimation of total organic carbon (TOC) content from well logs has become increasingly important because of the significance of TOC in the formation evaluation of those resources. In this paper, a new log overlay method is developed to estimate the TOC content of source rocks with excess radioactivity, but containing little or no potassium feldspar. Specifically, on the basis of previous results of log responses of source rocks, it is believed that the natural gamma ray (GR) log responses of source rocks in the applicable conditions are predominantly contributed by clay minerals and organic matter. A practical clay indicator is established to reflect the clay content using density and neutron logs. The indicator is effective not only in nonsource rocks that contain oil or water but also in source rocks. Furthermore, a new method was developed by overlaying the properly scaled clay indicator curve on the GR curve. In nonsource rocks, including clay-rich rocks and reservoirs saturated with oil or water, the two curves overlie each other, whereas a separation between the curves occurs in organic-rich source rocks. The separation between the curves was defined and expressed and can be used to calculate the TOC consecutively after careful calibration with core data. This method has been successfully applied to two shale gas plays with high-maturity kerogen in the Sichuan Basin, China. In addition, a source rock with low-maturity kerogen was used to verify the new method for its effectiveness, reliability, and widespread adaptability.

The wave power extraction by a cylindrical oscillating water column (OWC) device with a quadratic power take-off (PTO) model was studied experimentally and theoretically. In the experiment, a scaled model OWC was tested in a wave flume, with an orifice being used to simulate a quadratic PTO mechanism. In the theoretical analysis, the quadratic PTO model was linearized based on Lorenz's principle of equivalent work, which allows us to perform a frequency domain analysis using an eigen-function matching method. The effects of higher harmonic components and the spatial non-uniformity of the surface velocity inside the chamber were discussed. A semi-analytical model was proposed to understand the viscous loss affecting the measured capture length. Our treatment of the quadratic PTO model was validated by comparing quasi-linear theoretical capture length and the laboratory measurement. Our results also showed that the effects of spatial non-uniformity and viscous loss could be noticeable for shorter waves.

Potential security threats pose a significant challenge to evaluating the trustworthiness of complex links among users in a social network. Traditional trust computation methods typically consider user comments or interaction, thereby reflecting the trustworthiness between users according to their past experiences. However, the tie strength, which reflects the closeness of user relationships, is also a potential factor for estimating the trustworthiness of links among users. To incorporate this indicator, we propose a trust evaluation scheme for complex links comprising two aspects: the reliability and strength of links among uses. Our main contributions are (1) a trust calculation method, including direct trust for directly linked users and indirect trust for indirectly linked users, which is established based on the comment factor, forwarding factor, and approving factor; (2) a link strength evaluation method to determine the trustworthiness of direct and indirect links between users considering comment stability, mutual trust, interaction frequency, and common neighbours and community similarity, and (3) a link trust evaluation algorithm based on the link trust matrix synthesizing the reliability and strength of links. The experimental results and analysis show that our proposed scheme is feasible and effective in improving the performance of trust evaluation in a social network.

Given a set of k-dimensional objects, the skyline-CL query returns all clusters over skyline objects according to their cardinalities. A naive solution to this problem can be implemented in two phases: (1) using existing skyline query algorithms to obtain all skyline objects and (2) utilizing the DBSCAN algorithm to cluster these skyline objects. However, it is extremely inefficient in real applications because phases 1 and 2 are all CPU-sensitive. Motivated by the above facts, in this paper, we present Algorithm for Efficient Processing of the Skyline-CL Query (AEPSQ), an efficient sound and complete algorithm for returning all skyline clusters. During the process of obtaining skyline objects, the AEPSQ algorithm organizes these objects as a novel k-ary tree SI(k)-Tree which is first proposed in our paper, and employs several interesting properties of SI(k)-Tree to produce skyline clusters fast. Furthermore, we present detailed theoretical analyses and extensive experiments that demonstrate our algorithm is both efficient and effective.

Global warming caused by carbon emissions has been recognized as a challenge to human sustainable development, and low-carbon city development is widely considered as an effective strategy to address this challenge. Numerous emission reduction measures have been implemented, and considerable efforts have been devoted in promoting low-carbon city. This paper examines whether sufficient efforts have been paid to these typical emission sectors, including Building, Industry, Energy Transformation, and Transportation by referring to the shared responsibility of each sector. The shared responsibility of individual emission sector is calculated by applying energy consumption data in 2014 World Energy Balance. The efforts contributed in emission reduction by each sector are examined by analyzing the low-carbon city work plans of 24 representative sample cities, which are selected globally. The research results demonstrate that sufficient emission reduction efforts have been paid in the Building sector and Transportation sector. But the Industry sector and Energy Transformation sector are less-attended in addressing emission reduction. The reason for the sufficient efforts paid in the Building sector and Transportation sector is considered as that the efforts for emission reduction in these two sectors can bring more co-benefits. However, emission reduction in Industrial sector is generally considered to have the effects of holding back economic growth, and the emission reduction in the sector of Energy Transformation will need enormous investment for advanced technologies. Policy for emission reduction in the Industry sector and Energy Transformation sector is indispensable to promote low-carbon city. This study appeals that (1) low-carbon city can be effectively implemented only if carbon reduction policy is adopted to all industrial activities, (2) multiple channels of financial resources should be established to support cities to mitigate carbon emissions in Industry sector, (3) cooperation on the development of clean energy technology between cities should be promoted, and (4) efforts should be paid to reduce carbon emission from using traditional energy transformation equipment by improving their efficiency.

Social networks play important roles in information propagation and interactions among friends. Billions of information stream are created by users in Twitter and diffuse among relationship networks. Users may receive mass tweets posted by their followees every week. Thus, how to rank and display the tweets users are interested in and willing to interact with, is necessary to improve user's experience. Information propagation is usually driven by user interaction behaviors including retweet and comment. While these behaviors are mainly affected by the social influence and user interest, which dynamically change with time. In this paper, we propose a tweets ranking model to predict information diffusion based on dynamic social influence and personal interests. Our model combines these factors together by exploiting listwise algorithm such as ListNet. Experiments show that our model performs better than several baseline models, and both social influence and personal interests play important roles in tweets ranking.

Attention-based deep learning network models have shown obvious advantages on the sentence representation in many NLP tasks. While in the answer selection domain, applying attention-based deep learning model to capture complex semantic relations between question and answer is an extremely challenging task. In this paper, instead of simply using max-pooling in the pooling layer, we propose the two-level attentive pooling model which can efficiently select several key and high semantic-related matching words in the question-answer pair to improve the accuracy of answer selection. Specially, our model is built on top of the hybrid network which includes GRU and CNN to encode the complex sentence representation. The experimental evaluation on two popular datasets shows that our model has the good effectiveness and achieves the state-of-art performance in the answer selection task.

With the explosive growth of social network, the exploitation of social information in recommendation models has become increasingly significant. However, most existing models only made use of users’ social information, ignoring the value of items’ social information. Based on the above fact, we present a listwise learning to rank recommendation model, UIContextListRank, which generates a ranked list of items for individual users directly. We employ matrix factorization to construct a listwise objective function that measures the difference between the predicted lists and the real ones. Furthermore, we express users’ social contextual information as their trust friends and items’ social contextual information as their concurrent items, and incorporate the social contextual information of both users’ and items’ into the listwise model to improve recommendation quality. Moreover, we implement our proposed model in a distributed environment to tackle the challenge of overwhelming data. Experiments have been conducted on two real-world datasets to evaluate the proposed model. And the experimental results prove the model’s effectiveness and efficiency.

In this paper, we propose a real-time recommender system (PRACE) for taxi drivers to find a next passenger and start a new trip efficiently, based on historical GPS trajectories of taxis. To provide high-quality passenger-seeking advice, PRACE takes passenger prediction, road condition estimation, and earnings into ranking simultaneously. Different from many previous researchers, we not only pay more attention to the driving context of taxis (i.e., driving directions, positions, etc.) but also extract meaningful representations of these attributes, using deep neural networks. To enhance the effect of learning, the result of statistics is added to the input of models. Relying on the map meshing method, we treat the prediction task as a multi-classification problem rather than a regression problem and make comparisons with several state-of-the-art methods. Finally, we evaluate our method through extensive experiments, using GPS trajectories generated by more than 10,000 taxis from the same company over a period of two months. The results verify the effectiveness, efficiency, and availability of our recommender system.

Since the nonlinear properties of the autonomous land vehicles (ALVs) and the time-varying relationship between ego-vehicle and the desired path, it is difficult to tune the parameters of a path tracking controller for the autonomous driving of ALVs. Aiming at this problem, a novel learning based path tracking method is proposed in this paper, which is composed of the Stanley control structure and a learning based module. The input of the learning module is the relationship between current vehicle state and the desired path, and the learning output is the parameter k in the Stanley control structure. What we want to learn is to adaptive tune k according to current vehicle state. A near-optimal policy is obtained by neural dynamic programming (NDP), which is an online and model-free algorithm. The learning based module online tunes the parameter k of the Stanley control structure. The simulation results show that the proposed path tracking method possesses attractive performance.

Integrating wave energy converters with breakwaters is a promising concept for wave energy utilization. On the basis of fulfilling the wave protection demands, pile-supported Oscillating Water Column (OWC)-type breakwaters can also meet the local needs of electricity far from the lands. In the present study, the wave energy extraction and vortex-induced energy loss of pile-supported OWC-type breakwaters were analyzed based on a two-point measurement method. The importance of energy extraction and vortex-induced energy loss on the wave energy dissipation of pile-supported OWC-type breakwaters were experimentally investigated. It was found that the trends of energy extraction and vortex-induced energy loss were generally correlated. The effects of the pneumatic damping induced by top opening affected the vortex-induced energy loss more than the energy extraction. Results showed that a larger pneumatic damping was preferable for the purpose of increasing energy extraction, whereas for a smaller pneumatic damping the vortex-induced energy loss was more important to the energy dissipation. With increasing draft, the energy extraction decreased, but the vortex-induced energy loss complementally contributed to the total energy dissipation and made the energy dissipation at the same level as that for a shallower draft.

Due to complex and nonlinear dynamic model, speed tracking control with high precision and robustness is always a hard-deal task for autonomous vehicle control. In recent years, learning methods are gradually used to solve many complex control problems through the observed data from the real control systems. In this paper, a speed tracking controller is designed based on online continuous actor-critic learning. The input of the learning is the current vehicle state defined beforehand and the output is the actual brake or fuel command to track the desired speed. The online learning task is performed on a longitudinal model established based on the collected real vehicle data. An algorithm named continuous actor-critic learning Automaton(Cacla) is used for the policy learning. Simulation results illustrate that the designed controller is competent for tracking the expected speed with satisfied precision.

Wave absorbers are usually needed for reducing wave reflection from vertical walls. Different from traditional wave absorbers, which dissipate wave energy into unusable forms, an oscillating water column (OWC) structure can serve as a wave absorber with the potential to generate electricity. In this study, the hydrodynamic performance of a pile-supported OWC structure in front of a vertical wall was investigated experimentally. The surface elevations inside the OWC chamber were measured at two points for a better description of the motion of the OWC. It was demonstrated that the incident wave energy could be effectively attenuated by the OWC structure. Compared with the pile-supported slotted-barrier-type wave absorbers, the OWC-type wave absorbers could function as effectively as the slotted-barrier-type wave absorbers but with a potential to extract wave energy for electricity generation. The effects of the gap between the OWC structure and the vertical wall were examined, and the experimental r...

Post-translational modification of protein is one of the most important biological processions in the field of proteomics and bioinformatics. Pupylation is a novel post translational modification which the small, intrinsically disordered prokaryotic ubiquitin-like protein is conjugated to lysine residues of potential segments. Both the experimental and computational prediction methods of such modified sites have proved to be a challenging issue. Computational methods mainly aimed at extracting effective features from the potential protein segments. In this paper, the statistical feature of adjacent amino acid residues has been proposed and the novel feature is combined appearance of adjacent amino acid and the BLOSUM62 matrix. The Neural Network and the Naive Bayesian model have been employed as the classification model in this work. Such model will also be utilised to deal with many other issues in the field of computational biology.