Search

This article presents the design of an event-trigger-based networked robust dissipative tracking control systems. An event-triggered mechanism (ETM) is proposed by introducing the probability density distribution of communication delay. Thus, it can yield less conservative results. The tracking error system is transferred by a distributed communication delay system wherein the delay probability density is viewed as the kernel of the distributed delay. The threshold of the proposed ETM is designed as a dynamic parameter to adapt the transmission of the control signal. By applying the Lyapunov method, sufficient conditions are achieved to ensure the stability of the tracking error system with strictly dissipative performance, which includes $H_{\infty }$ performance and passive performance. Based on the dissipativity analysis condition, a co-design method with both tracking control and the ETM is derived in a united framework. Experiments on networked dc motor tracking system are provided to show the effectiveness of the presented method.

International audience; The Bloom filter, answering whether an item is in a set, has achieved great success in various fields, including networking, databases, and bioinformatics. However, the Bloom filter has two main shortcomings: no support of item deletion and no support of expansion. Existing solutions either support deletion at the cost of using additional memory, or support expansion at the cost of increasing the false positive rate and decreasing the query speed. Unlike existing solutions, we propose the Elastic Bloom filter (EBF) to address the two shortcomings simultaneously. Importantly, when EBF expands, the false positives decrease. Our key technique is Elastic Fingerprints, which dynamically absorb and release bits during compression and expansion. To support deletion, EBF can first delete the corresponding fingerprint and then update the corresponding bit in the Bloom filter. To support expansion, Elastic Fingerprints release bits and insert them to the Bloom filter. Our experimental results show that the Elastic Bloom filter significantly outperforms existing works.

Burst is a common pattern in data streams which is characterized by a sudden increase in terms of arrival rate followed by a sudden decrease. Burst detection has attracted extensive attention from the research community. In this paper, we propose a novel sketch, namely BurstSketch, to detect bursts accurately in real time. BurstSketch first uses the technique Running Track to select potential burst items efficiently, and then monitors the potential burst items and capture the key features of burst pattern by a technique called Snapshotting. Experimental results show that our sketch achieves a 1.75 times higher recall rate than the strawman solution.

Execution time, reliability, and energy consumption are the three main performance parameters of processor scheduling for real-time embedded systems. It is very meaningful to optimize the energy optimization of processor scheduling to satisfy the requirements on time limit and reliability. This paper tries to optimize the energy consumption of the processor under three constraints: the partial ordering relations between task modules, the time limit, and the reliability. Based on directed acrylic graph (DAG) and quantum particle swarm optimization (QPSO), two scheduling algorithms were developed for the problem, namely, DAG_QPSO_I and DAG_QPSO_II. The two algorithms were compared with each other, and with other intelligent algorithms. The results show that the proposed algorithms are superior in optimization effect and efficiency, DAG_QPSO_I makes energy consumption more efficient than DAG_QPSO_II, and DAG_QPSO_II meets stricter requirements on time limit and reliability than DAG_QPSO_I.

The "data pipe" model used by the existing Internet protocol stack is no longer ideal for many emerging applications, due to multimedia, multicast, mobility, machine learning, and network management challenges. A new learning-oriented network architecture is required to deal with these challenges and serve learning-centric applications in data centers, around network edges, and on mobile devices. This article focuses on the network for AI and AI for network for learning-oriented network architecture. This is done by leveraging, improving, and creating new learning techniques to determine and optimize protocol mechanisms and control policies. The new network architecture can provide ample research opportunities in network topology control, protocol design, and performance evaluation, aiming to network a truly dependable cyber-infrastructure. The learning-oriented network can also learn from applications and communications automatically and continuously while running on different infrastructures to support diverse requirements. In addition, the network can keep evolving its protocol mechanisms and control policies in an online manner. It does this while maintaining protocol security and preserving user privacy, to learn and perform more effectively and efficiently. Finally, the main challenges and opportunities of learning-oriented network are discussed, encouraging further research.

We address the event-based control of nonlinear cyber-physical systems subject to deception attacks. In particular, an improved Takagi–Sugeno (T–S) fuzzy model is employed to solve the mismatch problem between the fuzzy system and fuzzy controllers. From the perspective of attack detection, we construct a novel queuing model to depict deception attacks. Then, a switched event-based communication scheme is proposed, which dynamically converts with different attack modes. The idea is to appropriately reduce the number of triggers according to the severity level of attacks, which can further save network resources. By using piecewise Lyapunov functional methods, we find a solution to the co-design of fuzzy controllers and event-triggering parameters while the concerned system is guaranteed to be exponentially stable. Finally, we apply the proposed approach to a mass–spring–damper system, where the effectiveness is well verified.

This article investigates the problem of resilient control for the Takagi–Sugeno (T–S) fuzzy systems against bounded cyberattack. A novel memory-based event triggering mechanism (ETM) is developed, by which the past information of the physical process through the window function is utilized. Using such an ETM cannot only lead to a lower data-releasing rate but also reduce the occurrence of wrong triggering event. Furthermore, the frequency of event generation is relatively smoother than existing ETMs. From the current releasing instant to the next, two periods are designed. The ETM works only when the first period ends, thereby avoiding the Zeno behavior that commonly exists in continuous ETM designs. The control system is then formulated as a switched fuzzy control system with two modes in each releasing period. Based on an assumption of secure control, and the proposed ETM, sufficient conditions are obtained to guarantee the exponential stability of networked T–S fuzzy systems in the presence of deception attacks in secure sense. Finally, a single-link rigid robot is taken as an example to illustrate the advantages of theoretical results.

This paper mainly focuses on the problem of networked fault estimation (FE) and fault-tolerant control (FTC). A novel adaptive memory-based mechanism is proposed by introducing the latest piece of historical output information. The historical information at each instant is matched with a corresponding weight such that the closer information is, the more contribution to the releasing event. Many unexpected triggering events can be avoided under this communication protocol, especially for the scenarios of the system with jitter disturbance or random noise. Moreover, to make the instantaneous data releasing rate adapt to the requirement of the control system, a time-varying threshold of the event-triggered mechanism is designed. Therefore, the burden of network bandwidth can be greatly decreased. Based on this proposed communication protocol, a new fault and state estimation model is developed. The fault-tolerant controller uses the estimations to compensate for the influence induced by the network and the fault. Sufficient conditions are derived to co-design the parameters of FE, FTC, and adaptive memory-based event-triggered mechanism. Finally, the performance of the proposed communication mechanism, FE, and FTC is evaluated on an example of the F-404 engine system.

Conventional image retrieval techniques for Structure-from-Motion (SfM) suffer from the limit of effectively recognizing repetitive patterns and cannot guarantee to create just enough match pairs with high precision and high recall. In this paper, we present a novel retrieval method based on Graph Convolutional Network (GCN) to generate accurate pairwise matches without costly redundancy. We formulate image retrieval task as a node binary classification problem in graph data: a node is marked as positive if it shares the scene overlaps with the query image. The key idea is that we find that the local context in feature space around a query image contains rich information about the matchable relation between this image and its neighbors. By constructing a subgraph surrounding the query image as input data, we adopt a learnable GCN to exploit whether nodes in the subgraph have overlapping regions with the query photograph. Experiments demonstrate that our method performs remarkably well on the challenging dataset of highly ambiguous and duplicated scenes. Besides, compared with state-of-the-art matchable retrieval methods, the proposed approach significantly reduces useless attempted matches without sacrificing the accuracy and completeness of reconstruction.

The development of smartphones and social networks has brought great convenience to our lives. Due to the increasing requirements of user privacy, user data are protected by encryption protocol. Nevertheless, the encrypted traffic may still be identificated by a third party. In order to improve the privacy protection of users, it is necessary to study the existing encrypted user behavior system. The existing user behavior identification adopts the statistical features of encrypted traffic, which fluctuates greatly in different transmission environments. In this paper, we propose a Stable Features Identification Method(SFIM), which concentrate on filtering out the stable features from the encrypted traffic to identify user behavior. Based on the principle of maximum entropy, we put forward an approach to divide the distribution ranges of these stable features, and map the feature space into vector space. Our research focuses on multiple user behavior in the Instagram application. The best evaluation results achieve 99.8% accuracy, 99.3% precision, 99.3% recall, and 0.09% false positive rate(FPR) on average.

This brief studies the design of a memory-event-triggered $H_{\infty }$ filter for unmanned surface vehicles (USVs) with communication delays. To save precious communication resources, a memory-event-triggered scheme (METS) is constructed by utilizing the mean of historical outputs. A novel united framework of the filtering error system under the METS and communication delays is established as a distributed delay system. By using the Bessel-Legendre (BL) inequality to handle the distributed delay, sufficient conditions are derived for designing $H_{\infty }$ filters for USVs under METS. Finally, an illustrative example is simulated to show the advantage of the presented strategy.

The development of the Internet of Things (IoT) has brought various IoT services, which facilitate and enrich human life. All these services are at risk of privacy leakage. The privacy-preserving issue of IoT query, which is a typical service, has attracted much attention. A superior candidate for solving the above issue is classical cryptographic schemes based on the computational difficulty. With the advent of quantum computation, the security of such schemes may be broken by the strong ability of some advanced quantum algorithms. How to design a secure scheme for privacy-preserving query under the threat of quantum computation is crucial. To this end, we construct a general architecture of privacy-preserving query in IoT scenarios. Besides trust or honesty-but-curious models, we present a scheme for privacy-preserving query which is also valid in the scenario of untrusted devices. We provide a detailed security analysis. The result shows our scheme achieves private preservation of both service providers and clients.

This article focuses on $H_{\infty }$ load frequency control (LFC) of multi-area power systems with cyber-attacks and communication delays by a memory-event-triggered mechanism (METM). The dynamics of power system components, such as governor, turbine and power generator, are imitated by some circuit systems. To save the precious network bandwidth, a METM is constructed to decide which measured data should be transmitted as the control signal. In contrast to some memoryless ETMs, the proposed METM uses historic data over a fixed period. To guarantee no Zeno behavior, a positive constant time is waited after each trigger happens. By treating the exchanges of tie-line power between the $i$ th control area and other areas as exogenous disturbances, a novel decentralized distributed delay system is modeled to represent the decentralized $H_{\infty }$ LFC of a multi-area power system under the proposed METM with deception attacks and communication delays. By selecting a Lyapunov-Krasovskii functional (LKF) based on the distributed delay terms, new sufficient criterions are obtained to co-design the $H_{\infty }$ LFC controller and triggering parameters for power systems against deception attacks. Finally, the advantages of the presented method is shown via an illustrative case.

This article considers an event-triggered $H_\infty$ synchronization of neural networks (NNs) with mixed (discrete and distributed) delays. To release the communication burden, a novel weighted integral event-triggered scheme (IETS) is proposed based on the past information of the system dynamics. In this scheme, for the first time, a weighting function is proposed to weight the system state over a given period, which can be viewed as a forgetting factor. Moreover, a waiting time interval is added in the proposed IETS to exclude Zeno phenomenon. By constructing a novel Lyapunov–Krasovskii functional with the distributed delay kernel and the weighting function, sufficient linear matrix inequality conditions for the existence of an event-triggered controller that guarantees an exponential synchronization of the delayed NNs with an $H_{\infty }$ performance are derived. Finally, an illustrative example and an application to image encryption are used to demonstrate the advantage of the proposed approach.

This paper investigates the problem of memory-based event-triggered fuzzy control for cloud-aided active suspension systems (ASSs) against deception attacks. A novel memory-based event-triggered mechanism (ETM) which is sensitive to deception attacks is proposed. Compared to the general ETM, the system under the memory-based ETM has a higher average data releasing rate during deception attacks and external disturbance. Therefore, a better suspension performance of the cloud-aided ASS can be obtained. Meanwhile, the system without deception attacks can maintain a lower average data releasing rate, thereby reducing the occupation of the network resource. Moreover, such a memory-based ETM can mitigate the occurrence of wrong triggering event that is generated by some abrupt variation of the input of ETM. Sufficient conditions that guarantee the desired performance of cloud-aided ASSs are derived. Finally, an example of quarter-vehicle suspension system is provided to verify the effectiveness of the proposed method.

Mobile health systems predict health conditions based on multimodal signals. Users are often reluctant to provide their health status over privacy concerns. It is challenging to make health predictions without sufficient historical data from the users. In this paper, we propose a user-based collaborative filtering mobile health system. The system requests users to provide a few health labels. These labels are used to determine cohort similarity and discarded afterward to ensure privacy protection. The cohorts are designed to maximize user similarity across health labels, variable relationships, and sensor data. Our system predicts users based on the health information from their cohort. We empirically evaluate the system by conducting a ten-week longitudinal study to assess the health conditions of 212 hospital workers using mobile devices, wearables, and sensors. The results show successful cohort assignments with five health labels. Health predictions achieve promising performance without historical data. Our system demonstrates strong interpretability, predictability, and usability.

This study addresses the event-based control of networked switched systems subject to cyber attacks. A novel mode-based event-triggered mechanism is proposed, which is capable of characterising non-periodic denial-of-service (DoS) attacks. A switching control law is employed in each subsystem to weaken the negative effects caused by DoS attacks. The time sequences of system switchings, event triggerings and intermittent DoS attack behaviours are fully investigated, which yields a resultant closed control system. Then, by using piecewise Lyapunov functional methods, sufficient conditions are formulated to guarantee the concerned system exponentially stable. Meanwhile, the co-design methods for switching controllers and event-triggering parameters can be developed. Finally, a numerical simulation example and a practical example of a second-order oscillating circuit are presented to verify the proposed methods.

Mobile edge computing (MEC) is considered to be a promising paradigm to enable the new generation of wireless networks. Meanwhile, inspired by the great potential of non-orthogonal multiple access (NOMA) technique to improve the spectrum efficiency, this letter jointly considers these two technologies and studies a partial computation offloading scheme. Specifically, uplink NOMA is used in MEC where multiple mobile devices (MDs) partially offload their computing tasks to MEC server on the same time-frequency resource block, distinguishing in the power domain. We formulate a max-min computation efficiency (MMCE) problem to achieve the fair CE among MDs. To solve this non-convex problem, a heuristic search algorithm is adopted to obtain the optimal policy, including uplink transmit power allocation policy and local computing resource allocation policy. Simulation results show the superiority of the proposed scheme by contrast with several baseline schemes.

This paper investigates the event-triggered H∞ control for networked control systems under the denial-of-service (DoS) attacks. First, a novel system model is established considering random, time-constraint DoS attacks. Second, an event-triggered scheme including an off-time is proposed to reduce the unnecessary occupation of network resources, with which a prescribed minimum inter-triggering time is guaranteed and Zeno problem is avoided. Third, sufficient conditions for the existence of an event-triggered controller which ensures the exponential stability of the closed-loop system with desired H∞ performance are formulated in linear matrix inequalities (LMIs). Finally, the effectiveness of the proposed method is examined by two illustrative examples, where a real communication network based on the ZigBee protocol is utilized.

This article contributes to design an event-triggered $H_{\infty }$ controller for networked control systems with network channel delay. First, the network channel delay is modeled as a distributed delay with a probability density function as its kernel. Then, the closed-loop event-triggered control system is established as a distributed delay system. To make full use of the delay probability distribution, the Lyapunov–Krasovskii functional is constructed with the distributed kernel. By applying the Lyapunov method, sufficient conditions for ensuring the stability of the closed-loop system with prescribed $H_{\infty }$ performance are formulated in linear matrix inequalities. A numerical example shows that the proposed method is less conservative than some existing results considering delay distribution.

Affective states are associated with people's mental health status and have profound impact on daily life, thus unobtrusively understanding and estimating affects have been brought to the public attention. The pervasiveness of wearable sensors makes it possible to build automatic systems for affect tracking. However, constructing such systems is a challenging task due to the complexity of human behaviors. In this work, we focus on the problem of estimating daily self-reported affects from sensor-generated data. We first analyze the intra- and inter-subject differences of self-reported affect labels. Second, we explore different machine learning models as well as label transformation techniques to overcome the individual differences in self-reported responses estimation. We conceptualize three experimental settings including long-term and short-term estimation scenarios. Our experimental results show that the mixed effects model and label transformation can yield better estimation of individual daily affect. This work poses the basis for future sensor-based individualized and real-time affective digital and/or clinical interventions.

This paper is concerned with the co-design problem of event-triggered scheme and $H_{\infty }$ static output control of linear Markov jump systems with deception attacks. To save the previous communication resources, a mode-dependent event-triggered scheme is utilized based on system output. To describe the deception attacks, a random variable satisfying Bernoulli distribution is employed. By using a separation approach, sufficient linear matrix inequality conditions for the existence of event-triggered output controllers that ensure the stochastic stability with prescribed $H_{\infty }$ are obtained. Then, a co-design algorithm is proposed to obtain the trade-off between the communication cost and $H_{\infty }$ performance. Lastly, the validity of the developed method is verified by a numerical example and a practical single-link robot arm system.

Most of the widely used multi-view 3D reconstruction algorithms assume that object appearance is predominantly diffuse and full of good texture. For the objects that violate this restriction, the surface can hardly be reconstructed because such area lacks sufficient support from dense point clouds. To tackle this problem, we introduce a novel two-stage prior-guided method based on point clouds enhancement to enable the application of multi-view reconstruction approaches in such scenes. In the first stage, we optimize the original PlaneNet plane segmentation priors by taking advantage of the estimated depth map and confidence map from multi-view stereo. In the second stage, we correct and supply 3D point clouds for the weakly supported plane surface on the basis of the upgraded priors. Furthermore, we utilize a slight disturbance of the enhanced point clouds to facilitate the subsequent mesh reconstruction. The proposed point cloud enhancement approach is evaluated on the large-scale $DTU$ dataset. Our method significantly outperforms previous multi-view stereo state-of-the-arts. We also demonstrate weakly supported plane surface reconstruction results from real-world photos that are unachievable with either the methods aiming at preserving weakly supported surfaces or the traditional state-of-the-art 3D reconstruction systems.

This article studies the problem of nonfragile integral-based event-triggered control for uncertain cyber-physical systems under cyber-attacks. An integral-based event-triggered scheme is proposed to reduce the data transmissions and save the limited network resources. The triggering condition is related to the mean of system state over a finite time interval instead of instant system state. Random cyber-attacks in a communication channel are taken into account and described by a stochastic variable subject to Bernoulli distribution. A novel Lyapunov–Krasovskii functional based on Legendre polynomials is constructed, and the Bessel–Legendre inequality technique is employed to handle the integral term induced by the integral-based event-triggered scheme. Resorting to these treatments, sufficient conditions are established via a set of linear matrix inequalities to guarantee the asymptotic mean-square stability of the closed-loop system. Finally, a numerical example shows that the presented method is effective.

This paper presents the event-triggered control for Takagi–Sugeno (T–S) fuzzy networked systems with transmission delay. An integral-based model is proposed for designing a new event-triggered scheme, which relies on the mean of the system state and the last triggered state. To handle the asynchronous premises of the fuzzy system and fuzzy controller, a novel triggering condition is added into the event-triggered mechanism. Then, the closed-loop T–S fuzzy event-triggered control system is established as a distributed delay system. With the help of the Legendre polynomials and their properties, the co-design conditions of triggering parameters and controller gains are given in linear matrix inequalities to ensure the asymptotic stability of the resulting closed-loop system. Finally, an experiment via a practical wireless network is implemented to illustrate the effectiveness of the proposed approach.

Affective computing is broadly applied to decision making systems ranging from mental health assessment to employability evaluation. The heterogeneity of human behavioral data poses challenges for both model validity and fairness. The limited access to sensitive attributes (e,g., race, gender) in real-world settings makes it more difficult to mitigate the unfairness of the model outcomes. In this work, we focus on the heterogeneity of human behavioral signals and analyze its impact on model fairness. We design a novel method named multi-layer factor analysis to automatically identify the heterogeneity patterns in high-dimensional behavioral data and propose a framework to enhance fairness of behavioral modeling without accessing sensitive attributes.

At-scale evaluation of new data center network innovations is becoming increasingly intractable. This is true for testbeds, where few, if any, can afford a dedicated, full-scale replica of a data center. It is also true for simulations, which while originally designed for precisely this purpose, have struggled to cope with the size of today's networks. This paper presents an approach for quickly obtaining accurate performance estimates for large data center networks. Our system,MimicNet, provides users with the familiar abstraction of a packet-level simulation for a portion of the network while leveraging redundancy and recent advances in machine learning to quickly and accurately approximate portions of the network that are not directly visible. MimicNet can provide over two orders of magnitude speedup compared to regular simulation for a data center with thousands of servers. Even at this scale, MimicNet estimates of the tail FCT, throughput, and RTT are within 5% of the true results.

This article investigates the problem of memory-event-triggered H∞ output feedback control for neural networks with mixed delays (discrete and distributed delays). The probability density of the communication delay among neurons is modeled as the kernel of the distributed delay. To reduce network communication burden, a novel memory-event-triggered scheme (METS) using the historical system output is introduced to choose which data should be sent to the controller. Based on a constructed Lyapunov-Krasovskii functional (LKF) with the distributed delay kernel and a generalized integral inequality, new sufficient conditions are formed by linear matrix inequalities (LMIs) for designing an event-triggered H∞ controller. Finally, experiments based on a computer and a real wireless network are executed to confirm the validity of the developed method.

The steel industry is developing a design-by-advanced analysis specification for cold-formed steel construction. This effort provides an opportunity to utilize the latest nonlinear structural analysis (advanced analysis) to design steel structures based on their overall system behaviour. This paper concerns the system reliability calibrations of this design-by-analysis method, with a particular focus on cold-formed steel portal frames. Four typical portal frames are considered. The system reliability assessment takes into account all important random variables. A limit-state design criterion is developed which is consistent with a desired level of system safety.

This paper studies the event-triggered H ∞ static output feedback control of linear systems with unreliable communication. The unreliable phenomenon between the event-triggering unit and the controller is described by a stochastic variable with Bernoulli random binary distribution. To cast the considered problem in the robust control framework, the event-triggering scheme is presented by a time-delay form. A vertex structure separation strategy is utilized to handle control gain with interval variations, which could alleviate computation burden heavily. Resorting to a division of control gain from Lyapunov variable, a new method for the non-fragile H ∞ controller synthesis is established in the framework of linear matrix inequalities. Simulations are executed to show the validity of the proposed approach.

This paper is concerned with the reliable event-triggered H∞ output control of nonlinear systems with actuator faults. A dynamic triggering scheme depending on system outputs is implemented to reduce the amount of communication transmissions, which is different from existing constant triggering thresholds. The parameters of actuator faults are estimated via observer state. To compensate for the fault effects on systems, the reliable controller parameters are adjusted along with the obtained estimations. By using some technical lemmas, new sufficient conditions for the closed-loop system to be asymptotically stable with prescribed H∞ performance are formed in linear matrix inequalities. Lastly, simulations are implemented to demonstrate the validity of the proposed method.

We address the resilient event-based control of nonlinear cyber-physical systems subject to deception attacks. In particular, an improved Takagi–Sugeno (T-S) fuzzy model is employed to solve the mismatch problem between the fuzzy system and fuzzy controllers. From the attacker's point of view, we construct a novel queuing model to depict the intermittent behaviors of deception attacks. Then, a resilient event-based communication scheme is proposed, which is dynamically switched with different attack modes. The idea is to appropriately reduce the number of triggers according to the severity level of attacks, which can further save network resources. By using piecewise Lyapunov functional methods, we find a solution to the co-design of fuzzy controllers and event-triggering parameters while the concerned system is guaranteed to be exponentially stable. Finally, we apply the proposed approaches to a mass–spring–damping system, where the effectiveness is well verified.

Many information sources are not just sequences of distinguishable symbols but rather have invariances governed by alternative counting paradigms such as permutations, combinations, and partitions. We consider an entire classification of these invariances called the twelvefold way in enumerative combinatorics and develop a method to characterize lossless compression limits. Explicit computations for all twelve settings are carried out for i.i.d. uniform and Bernoulli distributions. Comparisons among settings provide quantitative insight., DCC 2021

Due to the high (human) cost of image annotation, lack of large-scale annotation data prevents computer vision models from fully solving image classification tasks. On the other hand, the Bert model [1] has shown that training large scale models with non-annotation text data is possible. Inspired by this, we propose a Self-supervised Visual Feature Learning and Classification framework (SVFLC) that can be applied to large scale training data without annotation. This approach is based on the contrastive predictive learning (CPL) method. By refining CPL using special data augmentation and new contrastive learning mechanisms, learning shape-biased features can be emphasized. In the next step, these features are used to produce pseudo labels via a clustering algorithm. Inspired by the recent research on noisy labels, we proceed to employ distance of the sample from cluster centers to eliminate low-confidence labels, and use soft triplet loss and classification loss jointly for achieving robust performance in the final classification. In this unsupervised learning paradigm, on the Imagenet dataset, our framework outperforms commonly used approaches. The unsupervised performance is lower than supervised and semi-supervised learning approaches, but our proposed framework is more suitable for general cases, and serves as a baseline algorithm for future improvement to the unsupervised learning paradigm.

In recent years, the campus social network mining has been widely concerned by scholars because it is closely related to the physical and mental health development of college students and the university management work. Compared with the typical social network mining tasks, the challenges of campus social network mining mainly lie in the difficulty of data acquisition and unlabeled data. Based on the phenomenon of homogeneity between friends, this paper proposed a social relationship mining method. Firstly, this paper carried out data dimension reduction operation for each student's daily consumption data, and then innovatively carried out the convolution operation for the consumption behavior data after dimension reduction, and finally obtained the characteristics of students' consumption behavior. The Multiple dimensional scaling algorithm is adopted for data dimension reduction, and the convolutional network is an improved Lenet-5 network. Then, the personalized ranking model is developed base on the phenomenon that the distance between friends is less than the distance between non-friends. To solve the problem of unlabeled data, we mined the friends' relationship from the student's library access control system swiping card records, and then taken the mined friends' relationship as the real friends' relationship to train the model. From the experimental results, the social network mining method proposed in this paper performs well.

Machine learning models are at the foundation of modern society. Accounts of unfair models penalizing subgroups of a population have been reported in domains including law enforcement, job screening, etc. Unfairness can spur from biases in the training data, as well as from class imbalance, i.e., when a sensitive group's data is not sufficiently represented. Under such settings, balancing techniques are commonly used to achieve better prediction performance, but their effects on model fairness are largely unknown. In this paper, we first illustrate the extent to which common balancing techniques exacerbate unfairness in real-world data. Then, we propose a new method, called fair class balancing, that allows to enhance model fairness without using any information about sensitive attributes. We show that our method can achieve accurate prediction performance while concurrently improving fairness.

The era of edge computing has arrived. Although the Internet is the backbone of edge computing, its true value lies at the intersection of gathering data from sensors and extracting meaningful information from the sensor data. We envision that in the near future, the majority of edge devices will be equipped with machine intelligence powered by deep learning. However, deep learning–based approaches require a large volume of high‐quality data to train and are very expensive in terms of computation, memory, and power consumption. In this chapter, we describe eight research challenges and promising opportunities at the intersection of computer systems, networking, and machine learning. Solving those challenges will enable resource‐limited edge devices to leverage the amazing capability of deep learning. We hope this chapter could inspire new research that will eventually lead to the realization of the vision of intelligent edge.

In this work, we target a known problem in representation learning that is: beyond coarse classification, how can we better model fine-grained categorization? To address this problem, we introduce Deep Subclass Linear Discriminant Analysis (DeepSDA), which utilizes intra-class variation and inter-class similarity during training. We could achieve multimodal classification by maximizing the ratio of between-subclass scatter matrix and within-subclass scatter matrix. We maximize the eigenvalues along the discriminative eignevector directions. Hence the deep neural network is able to learn more discriminative representation space and thus has higher class separation in the linearly separable latent space. We show that DeepSDA leads to significant improvements on diverse fine-grained categorization and attribute learning benchmarks.

The quality of aluminum alloy ingots is very important to the performance of products. But the casting process is complicated. Chemical composition and other factors also have a great impact on the quality of the ingots. In this paper, based on the actual production of an aluminum casting plant in China, we propose a data-driven ingot quality prediction method using XGBoost. This method includes four parts: data collection, database construction, data cleaning and relation model establishment. Among them, data collection and database construction are based on process knowledge. Make sure the data is realistic. At the stage of data cleaning, the features which have great influence on the quality of ingot casting but are neglected in production are found. In dealing with the missing value, the fitting method is adopted to solve the serious problem of missing industrial data. The ensemble learning method is used to build the relational model. It can completely reflect the aluminum alloy casting process and make accurate prediction of ingot quality quantitatively. As can be seen from a large number of comparative experiment, the XGBoost model works very well and gives the key factors that affect the quality of ingots.

This article investigates the event-triggered synchronization of delayed neural networks (NNs). A novel integral-based event-triggered scheme (IETS) is proposed where the integral of the system states, and past triggered data over a period of time are used. With the proposed IETS, the integral event-triggered synchronization problem becomes a distributed delay problem. Using the Bessel–Legendre inequalities, sufficient conditions for the existence of a controller that ensures asymptotic synchronization are provided in the form of linear matrix inequalities (LMIs). Illustrative examples are used to demonstrate the advantages of the proposed IETS method over other event-triggered scheme (ETS) methods. Moreover, this IETS method is applied to the image encryption and decryption. A novel encryption algorithm is proposed to enhance the quality of the encryption process.

We propose the width-resolution mutual learning method (MutualNet) to train a network that is executable at dynamic resource constraints to achieve adaptive accuracy-efficiency trade-offs at runtime. Our method trains a cohort of sub-networks with different widths (i.e., number of channels in a layer) using different input resolutions to mutually learn multi-scale representations for each sub-network. It achieves consistently better ImageNet top-1 accuracy over the state-of-the-art adaptive network US-Net under different computation constraints, and outperforms the best compound scaled MobileNet in EfficientNet by 1.5%. The superiority of our method is also validated on COCO object detection and instance segmentation as well as transfer learning. Surprisingly, the training strategy of MutualNet can also boost the performance of a single network, which substantially outperforms the powerful AutoAugmentation in both efficiency (GPU search hours: 15000 vs. 0) and accuracy (ImageNet: 77.6% vs. 78.6%). Code is available at https://github.com/taoyang1122/MutualNet.

The use of automatic methods, often referred to as Neural Architecture Search (NAS), in designing neural network architectures has recently drawn considerable attention. In this work, we present an efficient NAS approach, named HM- NAS, that generalizes existing weight sharing based NAS approaches. Existing weight sharing based NAS approaches still adopt hand-designed heuristics to generate architecture candidates. As a consequence, the space of architecture candidates is constrained in a subset of all possible architectures, making the architecture search results sub-optimal. HM-NAS addresses this limitation via two innovations. First, HM-NAS incorporates a multi-level architecture encoding scheme to enable searching for more flexible network architectures. Second, it discards the hand-designed heuristics and incorporates a hierarchical masking scheme that automatically learns and determines the optimal architecture. Compared to state-of-the-art weight sharing based approaches, HM-NAS is able to achieve better architecture search performance and competitive model evaluation accuracy. Without the constraint imposed by the hand-designed heuristics, our searched networks contain more flexible and meaningful architectures that existing weight sharing based NAS approaches are not able to discover., Comment: 9 pages, 6 figures, 6 tables. Nominated for ICCV 2019 Neural Architects Workshop Best Paper Award

The design-by-analysis approach lays the foundation for the next generation of structural design standards in which the strength and structural safety check are performed in a single step at system level, without recourse to a structural standard for individual member checks. However, to date, no attempt has been made towards considering the failure of joints in the structural analysis, without which the complete set of limit states cannot be accurately predicted. This paper presents a new analysis approach that incorporates macro-element joint models based on the Generalised Component Method in the FE analysis of steel frame buildings. The new analysis approach is termed Generalised Component Method-based finite-element (GCM-FE) analysis. The fundamental aspects and principles of the GCM-FE analysis approach are established in this paper, including the framework of GCM-FE analysis, the constitutive models for the connection components and the implementation of GCM-FE analysis in commercial numerical software including an automatic modelling technique. The GCM-FE joint modelling method is first validated against the experimental results of three steel beam-to-column connection types, including the bolted moment end-plate connection, top-and-seat angle connection and web angle connection. GCM-FE analysis is subsequently performed on a two-storey four-bay irregular steel frame, showing apparent advantages over the traditional analysis methods which adopt simplified joint models. The GCM-FE analysis not only provides the ultimate resistance and failure mode of the frame, but also accurately predicts the load-redistribution process inside the connections and the resultant effect on the structural framework.

Considering filter gains with additive interval variations, this study is devoted to the non-fragile H ∞ filtering of linear systems with an event-triggered transmission. To take into account the influence of these variations thoroughly, a structured vertex separator is adopted to alleviate the computation burden. Combining the Finsler lemma, conditions of designing the non-fragile filter to ensure the required H ∞ performance are formed by linear matrix inequalities. Compared to the norm-bounded method to handle the variations, the proposed method could provide less conservative results. Simulations are executed to show the validity of the proposed method.

Wearable sensors (smart watches, health/fitness trackers, etc.) are experiencing an explosion in popularity. Their pervasiveness allows for effective data collections to quantify human behavior in natural settings, enriching traditional behavioral science research opportunities. In this paper, we focus on the problem of affect estimation from sensor-generated data, whereas ground truth is available to us in the form of daily self-reported affective states. First, our analysis shows that individuals’ self-reported affect labels exhibit low variance, with a few samples deviating significantly from average observations, which may relate to experiencing out-of-the-ordinary events. Second, we explore different machine learning models as well as label transformation techniques to directly infer the self-reported scores from available sensor data. Experimental results show that mixed effects model and label transformation are able to better estimate the individual daily affect. This work poses the basis for future sensor-based individualized and real-time affective digital and/or clinical interventions.

Measuring and understanding self-reported behavioral responses plays an important role in various clinical applications. However, individual differences in physiology, perception, and behavior cause data heterogeneity in both selfreported responses and biosignals, which can pose challenges for machine learning models. Data heterogeneity not only impacts the performance of the models but can also yield biased predictions that may lead to unfair decision making. In this work, we will explore approaches to address the modeling challenges that rise from heterogeneity of human behavioral data.

Advanced nonlinear finite element analysis (FEA) enables the Direct Design Method (DDM) of a frame by computational analysis in place of designing members and connections to codified specific equations. The Direct Design Method has become a feasible design option and attracted increasing attention from the research community, designers and code writers. One of the research challenges of adopting the DDM is to develop a system-based limit state design criterion to ensure a stipulated level of structural reliability, similar to that of member-based design methods. The present paper derives the system resistance factors suitable for the DDM of typical spatial moment-resisting steel frames comprising rigid connections and members of compact sections under wind and gravity loads. The interaction of the system resistance factor and frame reliability is established, which can serve as a basis for the code writers to choose the appropriate system resistance factors for the DDM. The Australian Research Council

Cyberbullying is a major issue on online social platforms, and can have prolonged negative psychological impact on both the bullies and their targets. Users can be characterized by their involvement in cyberbullying according to different social roles including victim, bully, and victim supporter. In this work, we propose a social role detection framework to understand cyberbullying on online social platforms, and select a dataset that contains users’ records on both Instagram and Ask.fm as a case study. We refine the traditional victim-bully framework by constructing a victim-bully-supporter network on Instagram. These social roles are automatically identified via ego comment networks and linguistic cues of comments. Additionally, we analyze the consistency of users’ social role within Instagram and compare users’ behaviors on Ask.fm. Our analysis reveals the inconsistency of social roles both within and across platforms, which suggests social roles in cyberbullying are not invariant by conversation, person, or social platform.

Equipment management performance assessment is a multi-criteria decision-making process. In view of the randomness and fuzziness of the factors affecting the equipment management performance and the difficulties to evaluate objectively and quantitatively, this paper improves and perfects the model though weight determination and assessment method selection. To avoid the information loss and distortion, the combination weighting method is established and the cloud model is introduced into equipment management performance assessment. So this paper establishes an assessment model based on combination weighting and cloud model. The approach is tested by evaluating an example and comparing with fuzzy comprehensive evaluation method. The comparison verifies the feasibility and effectiveness of the algorithm.