Search

Recently, deep learning models have been extensively adopted in numerous applications, from health care to finance and entertainment industry. This wide-spread deployment of deep models raised concern over the privacy of data used to train deep models. This is a huge concern particularly for data-sensitive applications, such as health records, personal data, bio-metric data, etc. As a result, a new direction of research focusing on possible attacks aiming to identify training data of deep models emerged, called membership inference.Membership inference (MI) attacks identify which samples have been used during training and which samples have not. The first generation of membership inference attacks mainly used deep models' prediction confidence as a feature to identify training samples. The intuition is that deep models are more confident on samples they have seen during training than non-training samples. Despite their sound intuition and apparent successful reports, we, along a few other parallel studies, showed that the first generation of membership inference attacks are ineffective in practice for multiple reasons. First, they could not significantly outperform a naive baseline that labels a sample as a member (training sample) if it is correctly classified by the deep model and as a non-member (non-training sample) otherwise. Second, the confidence distribution of correctly classified samples, which cover the majority of a dataset, are not distinguishable between train and non-train samples. Only a small portion of mis-classified samples exhibit discrepant distribution. Third, all these membership inference attacks report average-case success metric (e.g., accuracy or ROC-AUC). However, privacy is not an average case-metric, and it should be treated similar to other security and privacy related problems. Similar to other security problems, the attack is reliable if it can identify a few training samples while almost on non-training samples are falsely labeled as a training sample. In other words, a reliable membership inference attack should have a decent true-positive rate (TPR) at low false-positive rates (FPR).In this dissertation, we aim to move the membership inference research in a more practical direction, either by showing the limitations of the current attacks or by proposing more reliable attacks. As stated earlier, we first show that the current generation of membership inference attacks are not reliable in practice. Then, we propose several new membership inference attacks that achieve more reliable performance in more realistic scenarios. The first attack focuses on the model's behavior in the entire sub-population, instead of a single sample in vacuum. More specifically, we compare the model's confidence on a target sample and other samples from the same sub-population. If the confidence of a sample is significantly higher than the average confidence on that sub-population, that is an indication of a training sample. We show that this attack can achieve moderate true positive with very low false positive. Additionally, we propose a BiGAN architecture to generate samples from the same sub-population, in case it is not available. The second attack aims to focus on user-level MI attack instead of the record-level MI attack. In this scenario, we identify if a user's data has been used during training instead if identifying which samples from the user have been used. Not only this attack is more realistic in privacy domain, but we show that we can achieve the state-of-the-art accuracy if multiple samples from a user are used to draw the membership inference. In another study, we show that MI attacks are generally more successful when deep ensemble is used. We show that deep ensemble shifts the distribution of train and non-train samples in a different way where they become significantly more distinguishable. Finally, we show that are a few simple aggregation mechanisms instead of ensemble averaging that can improve the accuracy and privacy of deep models in deep ensemble context.Finally, we illustrate a fundamental issue with current MI attacks, including the state-of-the-art attacks, that limits their applications in certain scenarios. We elaborate the issues with a practical scenario where membership inference attacks are used by an auditor (investigator) to prove to a judge/jury that the auditee unlawfully used sensitive data during training. Although the current SOTA attacks can identify some training samples with low false positive ratio in a common experimental setting extensively used for MI attacks, an auditee can generate unlimited number of samples on which MI attacks catastrophically fail. This can be used in court to easily discredit the allegation of the auditor and make the case dismissed. Interestingly, we show that auditee does not need to know anything about the auditor's membership inference attack to generate those challenging samples. We called this problem, discredibility. Currently, there is no attack immune to discredibility. We hope that our research sheds light on this newly-discovered issue and encourage researchers to investigate it.

In a real-world setting, object instances from new classes can be continuously encountered by object detectors. When existing object detectors are applied to such scenarios, their performance on old classes deteriorates significantly. A few efforts have been reported to address this limitation, all of which apply variants of knowledge distillation to avoid catastrophic forgetting. We note that although distillation helps to retain previous learning, it obstructs fast adaptability to new tasks, which is a critical requirement for incremental learning. In this pursuit, we propose a meta-learning approach that learns to reshape model gradients, such that information across incremental tasks is optimally shared. This ensures a seamless information transfer via a meta-learned gradient preconditioning that minimizes forgetting and maximizes knowledge transfer. In comparison to existing meta-learning methods, our approach is task-agnostic, allows incremental addition of new-classes and scales to high-capacity models for object detection. We evaluate our approach on a variety of incremental learning settings defined on PASCAL-VOC and MS COCO datasets, where our approach performs favourably well against state-of-the-art methods., Comment: Published in IEEE Transactions on Pattern Analysis & Machine Intelligence, Nov 2021. Code is available in https://github.com/JosephKJ/iOD

The social Internet of Things (IoT) (SIoT) helps to enable an autonomous interaction between the two architectures that have already been established: social networks and the IoT. SIoT also integrates the concepts of social networking and IoT into collaborative edge computing (CEC), the so-called CEC-based SIoT architecture. In closer proximity, IoT devices self-organize into a CEC-based SIoT computing cluster and provide social device-to-device (S-D2D) services, such as computation offloading, service discovery, and content delivery. In the CEC-based SIoT, however, cooperation based on social connections leads to a problem called social and spatial physical trade-off. This problem is also referred to as the mismatch problem, which arises because the spatial neighbors in the social layer cannot always be related. The spatial distance thus calls for additional multi-hop transmissions. This work presents a novel solution called 3-D-social identifier structure (3-D-SIS) model. The 3-D-SIS model is based on 3-D social space (3-D-SS) and considers social ties and physical connections (i.e., intra-neighbor) of the SIoT devices and utilizes a 3-D structure to evaluate that relationship. Moreover, it minimizes the end-to-end delay and communication cost to address the mismatch problem. To validate the performance of the (3-D-SIS) model, we use the real traces of social networks (INFOCOM06). The results show that the 3-D-SIS selects the best neighbor in S-D2D communication and improves performance in terms of end-to-end delay and throughput.

The Classical Model Predictive Control (CMPC) has the drawback of slow response in complex dynamic systems. In this work, the Fast Model Predictive Control (FMPC), which accelerates the computation time through the fragmental solution of a complex quadratic program (QP), is investigated as a possible alternative to control the standard CO2 capture plant using MEA with various step changes. Aspen PLUS® and MATLAB® a are utilized to implement the control strategy. The study concluded that the FMPC controller has an average settling time of 51.42 seconds for all step changes, which is 74.8% faster than the CMPC. The average IAE value for FMPC was approximately 0.1307 which is 59 times smaller than the CMPC controller. Additionally, the ISE and ITSE values demonstrated much improved outcomes for the FMPC controller. The offsets for the FMPC are maintained at negligible levels through suitable tunning since offsets are the main hurdle observed when the FMPC controller is implemented on chemical process systems.

Dissolved oxygen content prediction plays an important role in the intelligence management of aquaculture systems. Compared with traditional point prediction, interval prediction can quantify the uncertainties effectively and more closes to the fact. Unfortunately, there is rarely study about it. In this paper, a novel interval prediction method based on a deep auto-regression recurrent neural network (DeepAR) is proposed to construct prediction intervals (PIs) directly. Besides, a variational mode decomposition (VMD) has been used to extract the frequency feature of the original data. Moreover, a multi-objective weighted optimization framework based on the sparrow swarm algorithm (SSA) was proposed to improve PIs accuracy. Finally, simulations with water quality datasets were conducted to show the effectiveness of the proposed model. The results demonstrated that the proposed model significantly improved PI quality and performance compared to the state-of-the-art method.

This study proposes a Pade approximation based hybrid mesh free framework for numerical solutions of nonlinear partial differential equations. The proposed framework involves three novel aspects. Firstly, the conjunction of Pade approximation based residual functional and penalty function approach for handling boundary/Initial conditions is employed to design an equivalent optimization problem. Secondly, the notion of cooperative multi-simplex searches in the environment of Nelder-Mead Simplex (NMS) algorithm is introduced. Finally, a hybrid of multi-simplex search and Differential Evolution (DE) is proposed to efficiently cope with the modelled problem. The developed method is named as Pade Evolutionary Cooperative Multi-Simplex (PECMS) algorithm. The results of the developed PECMS scheme on illustrative examples are compared with three well-known methods, namely, Differential Transform Method (DTM), Adomian Decomposition Method (ADM) and the Variational Iteration Method (VIM). The comparisons demonstrate that the current approach possesses high accuracy when compared to the competing methods.

Disruptions in the manufacturing process cast substantial impacts on the consumer's centric values, cost, and overall efficiency. The ripple effect of these disruptions is observed in the whole ass...

Detecting an intruder that is trespassing a prohibited area is a critical task of intelligent surveillance systems. This task requires a change detector to segment an intruder (foreground object) from the background. The task suffers the inherent drawbacks of change detectors due to the dual-camera sensor (color/IR), illumination changes, night time, static, and camouflaged foreground objects. This article proposes an enhanced unsupervised change detector to compensate for the aforementioned challenges for industrial sterile zone monitoring. The camera switch detection based on skewness patterns detects a switch between the dual-camera sensors (color/IR). The optimal color space selection based on the mean squared error will select tolerant color space (RGB/YCbCr) to illumination changes for modeling the background. In addition, the IR camera frames are contrast-enhanced to tackle the camouflaged intruders during the night. The incoming frames are split into respective channels before modeling the background. The background is modeled by Gaussian mixture models. The adaptive background model update scheme is proposed to tackle the various challenges posed by environment and object such as a static foreground object. The comparison is performed on three databases with top-ranked unsupervised change detection algorithms.

The mean curvature-based image deblurring model is widely used to enhance the quality of the deblurred images. However, the discretization of the associated Euler-Lagrange equations produce a nonlinear ill-conditioned system which affect the convergence of the numerical algorithms like Krylov subspace methods. To overcome this difficulty, in this paper, we present two new symmetric positive definite (SPD) preconditioners. An efficient algorithm is presented for the mean curvature-based image deblurring problem which combines a fixed point iteration (FPI) with new preconditioned matrices to handle the nonlinearity and ill-conditioned nature of the large system. The eigenvalues analysis is also presented in the paper. Fast convergence has shown in the numerical results by using the proposed new preconditioners.

This research presents work on control of a prosthetic arm using surface electromyography (sEMG) signals acquired from triceps and biceps of fifteen healthy and four amputated subjects. Myo armband was used to acquire sEMG signals corresponding to four different arm motions: elbow extension, elbow flexion, wrist pronation, and wrist supination. Ten time-domain features were extracted and considered for classification to recognize the four-arm motions. These features and their various combinations were used to train four different classifiers, in both offline and real-time settings. It was found that the combination of signal mean and waveform length as a feature and k-nearest neighbors as classifier performed significantly better (p < 0.05) than all other combinations in both offline and real-time settings. The offline accuracies of 95.8% and 68.1% and real-time accuracies of 91.9% and 60.1% were obtained for healthy and amputated subjects, respectively. Results obtained using the presented scheme successfully demonstrate that using suitable features and classifier, classification accuracies can be significantly improved for transhumeral prosthesis, thereby, providing better, wearable and non-invasive control of prostheses using sEMG signals.

This article presents an approach to implement a Nonlinear Model Predictive Controller (NMPC) in real‐time with a non‐standard cost index. The proposed technique's applications are presented to maximize the energy produced by a Wave Energy Converter (WEC) when the cost index is a non‐quadratic piecewise discontinuous functional of some design variables. The presented framework is based on pseudo‐quadratisation and weight scheduling, which is implemented using the ACADO toolkit for MATLAB/Simulink. The proposed strategy features code generation and deployment on the real‐time target machines for industrial applications. The simulations and experiments confirm the success of the proposed approach in achieving the feasible operation of the NMPC and an optimal power capture by the wave energy converters.

Modern data centers run web-scale applications on tens of thousands of servers, generating tens of thousands of Remote Procedure Calls (RPCs) to backend services for each incoming user request. Tail latency, due to a small fraction of randomly slow RPCs, decreases the performance of these incoming requests, degrades users’ quality of experience, and limits disaggregation (applications’ ability to scale across a data center). We argue that current approaches to improve tail latency (especially, those bounding computation time) are insufficient, even with (reconfigurable-) hardware accelerators. Instead, to chop off the tail , datacenter services should dynamically trade correctness (or result quality) for timeliness, providing bounded latency with near-ideal accuracy. In this paper, we discuss how the increasing prevalence of machine learning (including search techniques like approximate nearest neighbor and PageRank), perceptual algorithms (like computational photography and image/video caching), and natural language processing lets modern hardware accelerators make these dynamic correctness tradeoffs while improving users’ quality of experience.

Camera-based surveillance systems largely perform an intrusion detection task for sensitive areas. The task may seem trivial but is quite challenging due to environmental changes and object behaviors such as those due to night-time, sunlight, IR camera, camouflage, and static foreground objects, etc. Convolutional neural network based algorithms have shown promise in dealing with these challenges. However, they are exclusively focused on accuracy. This article proposes an efficient supervised foreground detection (SFDNet) algorithm based on atrous deep spatial features. The features are extracted using atrous convolution kernels to enlarge the field-of-view of a kernel mask, thereby encoding rich context features without increasing the number of parameters. The network further benefits from a residual dense block strategy that mixes the mid and high-level features to retain the foreground information lost in low-resolution high-level features. The extracted features are expanded using a novel pyramid upsampling network. The feature maps are upsampled using bilinear interpolation and pass through a 3x3 convolutional kernel. The expanded feature maps are concatenated with the corresponding mid and low-level feature maps from an atrous feature extractor to further refine the expanded feature maps. The SFDNet showed better performance than high-ranked foreground detection algorithms on the three standard databases. The testing demo can be found at https://drive.google.com/file/d/1z_zEj9Yp7GZeM2gSIwYKvSzQlxMAiarw/view?usp=sharing .

Because of the consistently expanding demand of energy, the introduction of a decentralized microgrid based on energy resources will soon be the most exciting development in the power system. Microgrids (MGs), which are mainly based on inverters, are becoming more popular as they can handle different forms of renewable energy effectively. However, one of the most challenging areas of research is their control. In the last few years, many control strategies have been developed. In this review, different control methods have been discussed, applicable to the MG system. Furthermore, the comparative analysis of classical and modern control strategies is also considered. This survey guides the new researchers about all available control strategies and room for improvement towards the optimal solution of the MG control techniques. It also identifies several research gaps and future trends therein as well as provides a solution to manage problems in MGs. The strategies are then compared based on their applicability to different control requirements.

With the proliferation of the internet, new modes of access and dissemination have been invented. This paradigmatic shift is not only providing a stimulus for science and technology, but traditional fields of knowledge, such as religious studies and allied disciplines, are also among its beneficiaries. Once it was an uphill task to find a scholar well versed in Islamic inheritance law and ask for advice on the distribution of a deceased’s estate. Various online inheritance calculators have made it convenient, and to some extent, eliminated the inevitability of the consultation with scholars of inheritance. In this background, the paper analyzes some accessible online inheritance calculators and explores their strengths and weaknesses. For evaluative purposes, two benchmarks are devised: one is ‘accuracy score’ and another ‘efficiency grade.’ The outcome of the assessment is mixed. Despite the accessibility of online calculators, one should not repose outright confidence without being aware of their merits and demerits. Some calculators have evolved a smooth and efficient system to solve a large number of propositions of inheritance, while others lack proficiency and accuracy.

Unauthorized entrance in a prohibited area might create a security risk. An intelligent surveillance system should be able to mitigate such a problem by incorporating a sterile zone monitoring algorithm. The algorithm is challenged by a dual-camera sensors (color/IR), dynamic backgrounds, illumination changes, camouflaged, and static foreground objects, etc. This paper proposes an improved change detector (ICD) to mitigate the above-mentioned challenges. It employs a novel statistical decision criterion (SDC) based on skewness patterns. The SDC helps to differentiate time of day using the camera sensors (color/IR). The input frames are processed according to the time of day. For instance, IR input is image-enhanced to differentiate between camouflaged intruders from the background. Then input goes through Gaussian Mixture Models (GMM) based change detector to segment foreground (intruder). The foreground object is further cleansed using morphological operations for possible isolated noise and holes. The ICD was tested on three datasets and outperformed top-ranked change detection algorithms.

The tracking sentiment of the news entities over time provides important information to governments and enterprises during the decision-making process. Recently, it has attracted the attention of the research community as well due to its popularity in many applications including; tracking news about elections, e-commerce, and e-governance. However, most of the work is focused on English whereas limited contributions have been done for Arabic. Moreover, there are no annotated corpora in the Arabic news domain that can be used to perform the sentiment tracking task. In this research, we present an Arabic news corpus and its associated sentiment tracking system to monitor the sentiments towards news entities in the Arab world. Sentiment classification and Named Entity Recognition techniques are used to prepare the corpus for the tracking task. A sample dataset containing 7200 tweets was manually annotated to be used in building multiple classifiers and annotate more than 2.3M tweets using the semi-supervised technique. The results of sentiment classification by using different machine learning classifiers and internal testing set show that semi-automatically annotated dataset outperforms the manually annotated dataset by 23% and 16% on two-way and three-way classification respectively using F1-score. The tracking results illustrate that over time the sentiment tracking performs well at discovering the most popular entities, from social media and, tracking their shifts in different Arab regions. It can be used to detect the possible reasons for sentiment change over time and, to predict the future sentiment of the news entities.

Unmanned Aerial Vehicles (UAVs) have been recently used for different civilian applications such as remote sensing, search, and rescue (SAR), precision agriculture (PA), etc. A UAVs ability to sense and find targets remotely and, based on that, hover close to the target for a particular action makes it an ideal platform for the aforementioned applications. There has been extensive work carried out in the field of visual-based detection, navigation, and control, but the problem of detecting different ground targets and performing certain actions is still an open research area. This study proposes a novel framework for multiple target detection, recognition, and navigation of the UAV to the desired target and closely inspect it. This proposed framework can be deployed for accurately spot spraying in PA applications or SAR. The target detection and recognition in the framework are achieved through a computationally efficient Convolutional Neural Network (CNN) trained model, whereas the close inspection of the target is achieved through a PID-based tracking algorithm which ensures the UAV hover around the target for few seconds. The developed framework performed the desired objective in five stages employing Lawson’s control theory of sense, process, compare, decide and act. The target detection and recognition in the framework were validated with the field experiment, while the entire framework was validated through a variety of simulation flights conducted in Gazebo and PX4. The experiments’ results showed the versatility of the developed system to many complex missions where the targets are added or removed.

Controlling weed infestation through chemicals (herbicides & pesticides) is essential for crop yield. However, excessive use of these chemicals has caused severe agronomic and environmental problems. According to accurate weed detection, an appropriate dose of herbicides is recommended in site-specific weed management (SSWM) applications, which may ultimately promote chemical saving while enhancing its effectiveness. In this context of SSWM, an accurate detection and recognition system needs to be established for recognizing weeds and crops to carry out precise agrochemical treatments in real-time applications. Unmanned aerial vehicles (UAVs) and other robots offer potential in precision agriculture (PA) applications by monitoring farmland on a per-plant basis, as they have the capability of acquiring high-resolution imagery providing detailed information for the distribution of crops and weeds in the field. In this regard, UAVs offer a cost-efficient solution for providing excellent survey capabilities. In this study, a deep learning system is developed for identifying weeds and crops in croplands. The developed system was implemented and evaluated using high-resolution UAV imagery captured over two different target fields: pea and strawberry; the developed system was able to identify weeds with an average accuracy of 95.3%, whereas the overall average accuracy (crops and weeds) was 94.73% for both the fields. The average kappa coefficient of the developed system was 0.89. The developed deep learning system outperformed the existing machine learning and deep learning-based approaches on comparison and can be embedded into a precision sprayer for adopting the SSWM strategy.

Deep learning-based algorithms showed promising prospects in the computer vision domain. However, their deployment in real-time systems is challenging due to their computational complexity, high-end hardware prerequisites, and the amount of annotated data for training. This paper proposes an efficient foreground detection (EFDNet) algorithm based on deep spatial features extracted from an RGB input image using VGG-16 convolutional neural networks (CNN). The VGG-16 CNN is modified by concatenated residual (CR) blocks to learn better global contextual features and recover lost feature information due to several convolution operations. A new upsampling network is designed using bilinear interpolation sandwiched between $3\times 3$ convolutions to upsample and refine feature maps for pixel-wise prediction. This helps to propagate loss errors from the upsampling network during backpropagation. The experiments showed the effectiveness of the EFDNet in outperforming top-ranked foreground detection algorithms. EFDNet trains faster on low-end hardware and demonstrated promising results with a minimum of 50 training frames with binary ground-truth.

Billions of IoT devices and smart objects are already in operation today and even more are expected to be on the network over time. These IoT devices will generate enormous amounts of data that cannot be allowed to transmit on the network without end-to-end encryption or any trust and security mechanism. Currently, we have certificate authorities that certify the identity of a network device by binding its identity with its public key. However, these certificate authorities are centralized in structure and will not be able to individually certify billions of IoT devices entirely. In this paper, we propose that in an SDN-based IoT network, the identities, i.e., public keys and trust indices of IoT devices, can be stored on a blockchain to ensure immutability and tamper-resistance. The paper presents a novel scalable solution for key and trust management of IoT devices in IoT networks, with a successful proof-of-concept that proves the scalability of the proposed solution. The combination of an IoT network along with blockchain technology and software-defined networking (SDN) is effectively demonstrated through simulation that is able to store the public keys of IoT devices on the blockchain and route the network traffic efficiently through SDN. The performance of the proposed solution is evaluated in terms of throughput and access time delay. The results illustrate that access delay and throughput were not affected linearly or exponentially and the proposed solution shows no significant degradation in the performance with the increase in the number of IoT nodes and packets.

In this paper, we investigate two new Dombi aggregation operators on bipolar neutrosophic set namely bipolar neutrosophic Dombi prioritized weighted geometric aggregation (BNDPWGA) and bipolar neutrosophic Dombi prioritized ordered weighted geometric aggregation (BNDPOWGA) by means of Dombi t-norm (TN) and Dombi t-conorm (TCN). We discuss their properties along with proofs and multi-attribute decision making (MADM) methods in detail. New algorithms based on proposed models are presented to solve multi-attribute decision-making (MADM) problems. In contrast, with existing techniques a comparison analysis of proposed methods are also demonstrated to test their validity, accuracy and significance.

The human being leads to an improved lifestyle because of better healthcare facilities, which is possible by practicing “sharing.” Sharing of minimal, necessary Electronic Health Records (EHRs) with legitimate grounds have made it easier for researchers and scientists to apply statistics, develop medicines, and improve healthcare facilities. However, it invites many security and privacy concerns. The data may contain personally identifiable attributes; publishing such data will cause problems. Several studies have been proposed to maintain a balance between privacy and utility, but they were not up-to-the-mark. We propose an improved technique for anonymizing EHR, quasi-identifiers to retain data privacy, and maintains utility while keeping it anonymous. The proposed technique performs anonymization by assigning data into classes, adding uncertainty to it, based on the deviation. Based on experiments, it is concluded that the proposed scheme performs better in terms of privacy, utility and is better to its predecessor while making it difficult to trace back.

Generally, the modern networked energy systems are constructed from some coupled sub‐systems interconnected via their states and share their information through a communication network featuring time delays. This article suggests an efficient multi‐agent based cooperative‐distributed networked economic model predictive control scheme for energy consumption optimisation and reducing the cost of consumed electricity in network‐based smart energy systems considering communication network delays. In this regard, a buffer‐based moving horizon strategy with estimator is suggested to estimate the own states of every sub‐system and the coupled states, that is, the states in which their information exchange among sub‐systems and their values are not accessible due to communication time delays. Moreover, the boundedness of the estimation error as well as the stability of the closed‐loop system are established. The usability and proficiency of the suggested scheme are proved by applying the developed approach on two case studies.

Resource depletion is a concern for the global economy; many think that available resources on the planet will not be able to cater to an ever‐growing population. Thus, economies are trying to become circular, leaving behind the linear tradition linear approach. In the circular economy (CE), physical resources and energy are made to loop back into the supply chain (SC) for a more extended period. Proper selection of suppliers is an essential criterion towards proper execution of the CE principle in SC. In this research, we have constructed a framework for evaluating the supplier concerning the CE implementation. Further, this research identifies the criterion and sub‐criterion, which are pertinent for evaluating the supplier in CE context. Fuzzy‐based ‘Criteria Importance Through Inter‐Criteria Correlation (CRITIC)’ method is justifiably applied to determine the aggregated weights of the criteria. Finally, ‘Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)’ method is used to determine the suppliers' ranking in the Indian automobile industry. Six criteria and 24 sub‐criteria are obtained as per recent literature and then inputs from experts. ‘Environment’ criterion came out as the most favourable criterion with a subjective weight of 0.230. The current research is one of the first such attempts to provide criteria for supplier selection in a CE environment. The developed framework would help organisations in implementing CE‐based supplier selection. The identified criteria and sub‐criteria would provide organisations with means to evaluate suppliers and help suppliers develop an effective and efficient CE based on the SC.

Researches about semantic computing are usually evaluated by experiments, lacking support from the theory of computation. But traditional finite automata are defined on a fixed finite set of states. They will be invalid when a different application needs to change the states. However, semantic computing always asks for more robust automata that can modify itself according to the changes of states. Therefore, in this paper, we redefine traditional automata in more robust forms by semantic computing. The intuitive idea is when the state which an automaton is supposed to enter has changed, we drive automata into a state that is similar to the original one. We can find another similar state in the original state set to replace the changed state by semantic similarity. Based on the theories of semantic similarity, there always exists a similar state. Therefore, once a state has changed, we can always find a similar state to replace it, which means that we can empower automata to adapt to the changes of states. Our new automata bridge the gap between semantic computing and the theory of computation. Furthermore, we also redefine fuzzy finite automata and pushdown automata in more robust forms. Finally, we provide an application about the weather forecast, which indicates how to overcome the limitations of traditional automata.

Fault tolerant control systems can be used in the process machines such as Internal Combustion (IC) engines to achieve greater reliability and stability in the fault conditions. Thus, costly loss of production due to the unusual and unexpected shutdown of these machines can be avoided. The Air Fuel Ratio (AFR) control system is an important system in IC engines and faults in the sensors of this system will cause its shutdown creating costly production loss, therefore, fault tolerance is necessary for them. In this paper, an Active Fault Tolerant Control System (AFTCS) based on Artificial Neural Networks (ANN) has been proposed for the AFR control system of a Spark Ignition (SI) IC engine to increase its reliability. In the proposed AFTCS, a nonlinear ANN-based observer is used in the Fault Detection and Isolation (FDI) unit for the highly nonlinear sensors of the AFR system for analytical redundancy. The Lyapunov stability analysis has been utilized to design a stable system in normal and faulty conditions. The system has been implemented in MATLAB/Simulink environment to test its performance. The simulation experimental results demonstrate that the suggested system stays reliable maintaining the stability well in the fault conditions of sensors with little degradation in AFR. A comparison with the existing works demonstrates the superior performance of the proposed AFTCS for the highly nonlinear sensors of the AFR control system. The technique suggested is very effective in terms of fault robustness and is more specifically based on the nonlinear behavior of the MAP sensor compared to the existing works.

The high contents of CO2 in natural gas processing industries cause various issues in the operation, and it is essential to reduce the amount of CO2 using amine-based absorption processes. An efficient and flexible control strategy is highly desirable for CO2 removal in CO2 absorption/stripping system. The excellent performance of Model Predictive Control (MPC) in setpoint tracking and disturbance rejection scenarios could make it a better option for the flexible controllability of a CO2 absorption/stripping plant. MPC performance depends significantly on the accuracy of the identified mathematical model of the plant. The state-space model is believed as the better option for MPC as it represents the plant model with true dynamics. Therefore, this study is focused on the design of the 2 × 2 MPC control strategy in MATLAB® MPC Designer Toolbox using 2nd order continuous-time state-space model. The main aim of this study is to develop a plant-wide control scheme based on MPC for the natural gas absorption/striping system. Step changes in the CO2 composition of sweet gas (±5%) and stripper temperature (±15%) have been introduced in the absorption/stripping simulation model. The results show that the MPC controller has achieved the new setpoint of CO2 composition within 0.5 sec in the setpoint tracking scenario. Similarly, the MPC controller has been able to reject the disturbances successfully introduced as ± 15% step change in stripper temperature within 7.5 sec. Hence, the performance of the MPC controller using the state-space model at higher step changes is adequate with no peak, closer to the setpoint, and no overshoot in the output.

As multi-hop wireless networks are attracting more attention, the need to evaluate their performance becomes essential. In order to evaluate the performance metrics of multi-hop wireless networks, including sending and receiving rates of a node as well as the collision probability, a model based on Stochastic Reward Nets (SRNs) is proposed. The proposed SRN models a typical node in such networks, considered as a general template to be applied to any wireless node. The SRN model of a single node is designed to take transmission effects of all neighboring nodes into account, while ignoring the ones whose transmission has no effect on the node under-study. Applying the proposed SRN to each node in a multi-hop wireless network, we can model any arbitrary topology and evaluate the mean queue size, collision probability, sending and receiving rates of the nodes. Common issues related to Medium Access Control (MAC) and physical layer of wireless networks, e.g. hidden and exposed node problems, collision, transmission, interference and carrier sense ranges, are considered in the proposed SRN model. The results obtained by the proposed model, in two different scenarios, are validated with the results obtained from the discrete-event simulation conducted in ns-2 simulator to show the applicability and accuracy of the proposed SRN model.

This paper presents new and efficient modulation techniques applied on the recently introduced compact nine-level switched-capacitor inverter (C9LSCI). The paper also discusses the performance of the inverter under selective harmonic elimination (SHE) and mitigation (SHM) based on the heuristic Bat-Algorithm (BA) technique. Two new modulation techniques employing CD-type carrier waves and M-type carrier waves are proposed, showing a rise in efficiency, an increase in RMS voltage, reduction in capacitor voltage ripple, and a reduction in harmonic distortion (THD). Four recently introduced modulating signals, namely the third harmonic injection, thirty-degree bus clamped PWM (THTDBCPWM), third harmonic injection sixty-degree bus clamped PWM (THSDBCPWM), one pole clamped PWM (OPCPWM). Loss balancing two-pole clamped PWM (LBTPCPWM) are applied to the inverter with the proposed carriers, and a comparative analysis is carried out on different performance parameters. The proposed carriers’ efficacy in Alternate Phase Opposition Disposition (APOD), Phase Opposition Disposition (POD), and Phase Disposition (PD) is validated. Further, Bat algorithm (BA) is applied as a heuristic approach to determine the optimum angles for the application of SHE and SHM. The efficacy of all the above techniques on the C9LSCI is verified in MATLAB/Simulink environment and further validated on an experimental setup.

Limited non-invasive transhumeral prosthesis control exists due to the absence of signal sources on amputee residual muscles. This paper introduces a hybrid brain-machine interface (hBMI) that integrates surface electromyography (sEMG) and functional near-infrared spectroscopy (fNIRS) signals to overcome the limits of an existing myoelectric upper-limb prosthesis. This hybridization aims to improve classification accuracy (CA) to escalate arm movements’ control performance for individuals who have a transhumeral amputation. To evaluate the effectiveness of this hBMI, fifteen healthy and three transhumeral amputee subjects for six arm motions were participating in the experiment. Myo armband was used to acquire sEMG signals corresponding to four arm motions: elbow extension, elbow flexion, wrist pronation, and wrist supination. Whereas, fNIRS brain imaging modality was used to monitor cortical hemodynamics response from the prefrontal cortex region for two hand motions: hand open and hand close. The average accuracy of 94.6 % and 74% was achieved for elbow and wrist motions by sEMG for healthy and amputated subjects, respectively. Simultaneously, the fNIRS modality showed an average accuracy of 96.9% and 94.5% for hand motions of healthy and amputated subjects. This study demonstrates the feasibility of hybridizing sEMG and fNIRS signals to improve the CA for transhumeral amputees, improving the control performances of multifunctional upper-limb prostheses.

Pedestrian detection relying on deep convolution neural networks has made significant progress. Though promising results have been achieved on standard pedestrians, the performance on heavily occluded pedestrians remains far from satisfactory. The main culprits are intra-class occlusions involving other pedestrians and inter-class occlusions caused by other objects, such as cars and bicycles. These result in a multitude of occlusion patterns. We propose an approach for occluded pedestrian detection with the following contributions. First, we introduce a novel mask-guided attention network that fits naturally into popular pedestrian detection pipelines. Our attention network emphasizes on visible pedestrian regions while suppressing the occluded ones by modulating full body features. Second, we propose the occlusion-sensitive hard example mining method and occlusion-sensitive loss that mines hard samples according to the occlusion level and assigns higher weights to the detection errors occurring at highly occluded pedestrians. Third, we empirically demonstrate that weak box-based segmentation annotations provide reasonable approximation to their dense pixel-wise counterparts. Experiments are performed on CityPersons, Caltech and ETH datasets. Our approach sets a new state-of-the-art on all three datasets. Our approach obtains an absolute gain of 10.3% in log-average miss rate, compared with the best reported results on the heavily occluded HO pedestrian set of the CityPersons test set. Code and models are available at: https://github.com/Leotju/MGAN .

Blind source separation (BSS) is a problem that appears in many research fields. Fast Independent components analysis (FastICA) is one of the techniques to solve the problem. The researchers have verified the effectiveness of the technique through the offline analysis of the public datasets. The development of a real-time portable system involving such a computationally complex analysis requires an efficient hardware implementation of FastICA. A Field programmable gate array (FPGA) and an application-specific integrated circuit (ASIC) are two promising hardware platforms to implement FastICA. This work proposes a new method, called ALgebraic Jacobi Method (ALJM), for performing eigenvalue decomposition (EVD) required for the implementation of FastICA. We use a simplification, a polynomial approximation, and the Newton-Raphson method for calculating the Jacobi rotation. In this way, we ensure hardware reusability between the EVD stage and the weight vector estimation (WVE) stage of FastICA which reduces the computational complexity and the power consumption, without compromising its computation speed. We evaluate the ALJM-based FastICA by performing BSS on the linear mixtures of the deterministic and the random signals and comparing the performance results with the existing methods. After verifying its functionality and numerical stability, we propose a scalable systolic processing array (SPA) for the ALJM-based FastICA and implement it on Spartan-6 FPGA. By comparing the existing implementations of FastICA, in terms of speed, area, and power, we conclude that the ALJM-based FastICA is one of the most efficient methods for prototyping and commercializing a real-time portable system comprising FastICA.

Elevators are surfaces of flight control, typically at the rear of an aircraft to control the pitch of the plane, the angle of attack and the wing lift. The most critical actuation device is longitudinal aircraft control, and its failures will result in a catastrophic aircraft crash. This paper proposes a Highly Redundant Fault Tolerant Control (HRFTC) policy for the aircraft to accommodate faults in the critical sensors and actuators. Modified Triple Modular Redundancy (MTMR) has been proposed for the sensors and Dual Redundancy (DR) has been proposed for the actuators. The working of control laws, pilot order, signal conditioning, and failure are elaborated. Furthermore, the PID controller is used for the adjustment of the position of the elevator by comparing it with a set point. The results show that when a fault occurs, the system detects it successfully and tolerates it quickly without disturbing the flight of aircraft. The study is significant for the avionics industry for manufacturing highly reliable machines for human and environmental safety.

Realizing an accurate laying rate prediction based on environmental factors plays a vital role in livestock and poultry breeding. In this paper, multiple environmental factors were considered to improve the accuracy of egg production rate prediction. A method was proposed by combining the Random Forest (RF) and Long Short-Term Memory (LSTM) to analyze the impact of the external environmental factors on the laying rate. Firstly, using RF, feature importance selection was implemented on environmental factors affecting laying rate. Secondly, the extreme Gradient Boosting (XGBoost) was introduced as a comparison to evaluate the accuracy and reliability of the RF feature importance selection. Finally, by discarding the features with low importance one by one, the multi-variable RF-LSTM laying rate prediction was conducted. Experiment results showed that the proposed RF-LSTM method significantly improved the prediction accuracy on laying rate.

Smartphones have an essential role in day-to-day lives. Mobile phone applications have several and wide ranging features. The use of these smartphone applications provides new insights for innovation and opportunities in language teaching and learning process. The present study intended to examine the efficiency of a mobile application on teaching 40 phrases from Nation and Martinez phrase test 1-5 compared to conventional activities. Quasi-experimental research design with pre-test and post-test was employed to determine the differences between the scores of the control group (N=20) and the experimental group (N=20) developed by using random sampling. Results indicated that participants in the experimental group performed significantly better in the post-test, demonstrating the effectiveness of the mobile application used in this study on learning idioms. The results showed that learners in the experimental group achieved considerably better development in the post-test, representing the efficiency of the smartphone application that was involved in the study. Suggestions have been given for using mobile phone applications more effectively for vocabulary learning.

Weeds affects crops health as it shares water and nutrients from the soil, as a result it decreases crop yield. Manual weedicide spray through bag-pack is hazardous to human health. Localized autonomous weedicide spray through aerial spraying units can help save water, weedicide chemical and effect less on human health. Such systems require multi-spectral cues to classify crop, weed, and soil surface. Our focus in this paper is on the detection of weeds in the sugar beet crop, using air-borne multispectral camera sensors, which is considered as an alternative crop to sugarcane to obtain sugar in Pakistan. We developed a new framework for weed identification; a patch-based classification approach as appose to semantic segmentation that is more realistic for real-time intelligent aerial spraying systems. Our approach converts 3-class pixel classification problem into a 2-class crop-weed patch classification problem which in turns improves crop and weed classification accuracy. For classification, we developed a new VGG-Beet convolutional neural network (CNN), which is based on generic VGG16 (visual graphics group) CNN model with 11 convolutional layers. For experiments, we captured a sugar beet dataset with 3-channel multispectral sensor with a ground sampling distance (GSD) of 0.2 cm/pixel and a height of 4 meters. For better comparison, we used two publicly available sugar beet crop aerial imagery datasets, captured using a 5-channel multispectral sensor and a 4-Channel multispectral sensor with a ground sampling distance of 1cm and a height of 10 meters. We observed that patch-based method is more robust to different lighting conditions. To produce low cost weed detection system usage of Agrocam sensor is recommended, for higher accuracy Red Edge and Sequoia multispectral sensors with more channels should be deployed. We observed higher crop-weed accuracy and lower testing time for our patch-based approach as compared to U-Net and Deeplab based semantic segmentation networks.

As the world is going through an existential global health crisis, i.e., the outbreak of novel coronavirus-caused respiratory disease (Covid-19), the healthcare systems of all the countries require readily available, low cost and highly precise equipment for the rapid diagnostics, monitoring, and treatment of the disease. The performance and precision of this equipment are solely dependent on the sensors being used. The advancement in research and development of micro-electro-mechanical systems (MEMS) based sensors during recent years, has resulted in the improvement of the conventional equipment being used in biomedical and health care applications. Microfluidics (Lab-on-a-chip) and MEMS sensors are now being used extensively for quick and accurate detection, progression monitoring, and treatment of various diseases including Covid-19. The ongoing miniaturization and design improvements have resulted in more precise sensors and actuators for healthcare applications, even for micro and nanoscale measurements in drug delivery and other invasive applications. This article aims at reviewing the MEMS sensors being used or which can be used in the important equipment for the detection and treatment of Covid-19 or other pandemics. An insight into various designs and working principles of the research-based and commercially available MEMS sensors is presented. The study highlights the role and importance of MEMS sensors in the improvement of equipment with conventional sensors. MEMS sensors outperform the conventional sensors due to their small size ( $1~\mu \text{m}$ -1mm), negligible weight, prompt response, precise measurements, portability, and ease of integration with electronic circuitry.

This paper presents comparative evaluation of an application of deep convolutional neural networks (dCNNs) to age invariant face recognition. To this end, we use four distinct dCNN models, the AlexNet, VGGNet, GoogLeNet and ResNet. We assess their performance to recognize face images across aging variations, firstly by fine-tuning the models and secondly using them as face feature extractor. We also suggest a novel synthesized aging augmentation technique suitable for age-invariant face recognition using dCNNs. The face recognition experiments are conducted on three challenging FG-NET, MORPH and LAG aging datasets, and results are benchmarked with a simple CNN. The comparative study allows us to answer (i) when and why transfer learning or feature extraction strategies are useful in age-invariant face recognition scenarios, (ii) the potential of aging synthesized augmentation to increase accuracy and (iii) the choice of appropriate feature normalization and distance metrics to be used with deeply learned features. The extensive experiments, and valuable insights presented in this study can be extended to the design of effective age-invariant face recognition algorithms.

For the modern power systems with intensive renewable energy sources integration, the uncertainty analysis of its performance become necessary for desired decision making. An appropriate choice of ...

Routing is one of the major challenges in wireless sensor networks (WSNs). Unbalanced energy consumption in the routing process of data packets is one of the main issues in WSNs. The issue needs consideration, because the energy level of sensor nodes is limited. Multipath routing methods reduce energy consumption, improve scalability and provide load balancing in WSNs. In this study, we suggested a multipath routing method for homogeneous WSNs. The proposed method includes 3 phases: clustering the network nodes, discovering the paths between CHs, and maintaining the paths. In the first phase, wireless sensor network is clustered through the firefly algorithm. In the second phase, routing is performed between CHs based on the fuzzy logic. Routing between CHs results in creating 2 paths: primary path and backup path. CHs transmit data packets to the base station through the primary paths; however, failures in primary paths cause CHs to employ backup paths. In the third phase, the paths are maintained so that path breakages cause to restart route discovery. The results of the simulation reveal that the proposed multipath routing outperforms other routing methods in end-to-end delay, energy consumption, packet loss rate, and network lifetime.