Search

The accurate and efficient prediction of chlorophyll-a (Chl-a) concentration is crucial for the early detection of algal blooms in reservoirs. Nevertheless, predicting Chl-a concentration in multivariate time series poses a significant challenge due to the complex interrelationships within the aquatic environment and the discrete and non-stationary nature of online monitoring of water quality data. To address the aforementioned issue, this paper proposes a novel prediction model named SGMD-KPCA-BiLSTM (SKB) for predicting Chl-a concentration. The model combines two-stage data processing and machine learning (ML). To capture nonlinear relationships in multivariate time series data, the optimal data subset is determined by combining symplectic geometry mode decomposition (SGMD) and kernel principal component analysis (KPCA). This subset is then input into a bidirectional long short-term memory (BiLSTM) model, and the model's hyperparameters are optimized using the sparrow search algorithm (SSA) to improve the accuracy of predictions. The performance of the model was evaluated at Qiaodian Reservoir in Shandong, China. To assess its superiority, the evaluation criteria included the root mean square error (RMSE), mean absolute percentage error (MAPE), mean absolute error (MAE), coefficient of determination (R 2 ), frequency histograms of the prediction error, and the Taylor diagram. The prediction performance of five single models, namely the back-propagation (BP) neural network, support vector regression (SVR), long short-term memory (LSTM), convolutional neural network with long short-term memory (CNN-LSTM), and BiLSTM, as well as three hybrid models, namely SGMD-LSTM, SGMD-KPCA-LSTM, and SGMD-BiLSTM, were compared against the SKB model. The results demonstrated that the SKB model performs best in predicting Chl-a concentration (R 2  = 96.19%, RMSE = 1.05, MAE = 0.65, MAPE = 0.08). It significantly reduced the prediction error compared to other models for comparison. Furthermore, the multi-step predictive capabilities of the SKB model are also discussed. The analysis shows a decline in predictive performance with larger prediction time steps, and the SKB model exhibits slightly superior performance compared to the other model at corresponding prediction intervals. The model has significant advantages in terms of its ability to accurately predict the non-smooth and nonlinear Chl-a sequences observed by the online monitoring system. This study presents a potential solution for controlling and preventing reservoir eutrophication, as well as an innovative approach for predicting water quality., (© 2023. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

This study presents a novel method to assess the learning effectiveness using Electroencephalography (EEG)-based deep learning model. It is difficult to assess the learning effectiveness of professional courses in cultivating students' ability objectively by questionnaire or other assessment methods. Research in the field of brain has shown that innovation ability can be reflected from cognitive ability which can be embodied by EEG signal features. Three navigation tasks of increasing cognitive difficulty were designed and a total of 41 subjects participated in the experiment. For the classification and tracking of the subjects' EEG signals, a convolutional neural network (CNN)-based Multi-Time Scale Spatiotemporal Compound Model (MTSC) is proposed in this paper to extract and classify the features of the subjects' EEG signals. Furthermore, Spiking neural networks (SNN) -based NeuCube is used to assess the learning effectiveness and demonstrate cognitive processes, acknowledging that NeuCube is an excellent method to display the spatiotemporal differences between spikes emitted by neurons. The results of the classification experiment show that the cognitive training traces of different students in solving three navigational problems can be effectively distinguished. More importantly, new information about navigation is revealed through the analysis of feature vector visualization and model dynamics. This work provides a foundation for future research on cognitive navigation and the training of students' navigational skills.

Deep convolutional neural networks (DCNNs) have demonstrated promising performance in classifying breast lesions in 2D ultrasound (US) images. Exiting approaches typically use pre-trained models based on architectures designed for natural images with transfer learning. Fewer attempts have been made to design customized architectures specifically for this purpose. This paper presents a comprehensive evaluation on transfer learning based solutions and automatically designed networks, analyzing the accuracy and robustness of different recognition models in three folds. First, we develop six different DCNN models (BNet, GNet, SqNet, DsNet, RsNet, IncReNet) based on transfer learning. Second, we adapt the Bayesian optimization method to optimize a CNN network (BONet) for classifying breast lesions. A retrospective dataset of 3034 US images collected from various hospitals is then used for evaluation. Extensive tests show that the BONet outperforms other models, exhibiting higher accuracy (83.33%), lower generalization gap (1.85%), shorter training time (66 min), and less model complexity (approximately 0.5 million weight parameters). We also compare the diagnostic performance of all models against that by three experienced radiologists. Finally, we explore the use of saliency maps to explain the classification decisions made by different models. Our investigation shows that saliency maps can assist in comprehending the classification decisions., (© 2023. The Author(s).)

XAI is a rapidly progressing field of machine learning, aiming to unravel the predictions of complex models. XAI is especially required in sensitive applications, e.g. in health care, when diagnosis, recommendations and treatment choices might rely on the decisions made by artificial intelligence systems. AI approaches have become widely used in aging research as well, in particular, in developing biological clock models and identifying biomarkers of aging and age-related diseases. However, the potential of XAI here awaits to be fully appreciated. We discuss the application of XAI for developing the "aging clocks" and present a comprehensive analysis of the literature categorized by the focus on particular physiological systems., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier B.V. All rights reserved.)

Artificial intelligence (AI) is often used to describe the automation of complex tasks that we would attribute intelligence to. Machine learning (ML) is commonly understood as a set of methods used to develop an AI. Both have seen a recent boom in usage, both in scientific and commercial fields. For the scientific community, ML can solve bottle necks created by complex, multi-dimensional data generated, for example, by functional brain imaging or *omics approaches. ML can here identify patterns that could not have been found using traditional statistic approaches. However, ML comes with serious limitations that need to be kept in mind: their tendency to optimise solutions for the input data means it is of crucial importance to externally validate any findings before considering them more than a hypothesis. Their black-box nature implies that their decisions usually cannot be understood, which renders their use in medical decision making problematic and can lead to ethical issues. Here, we present an introduction for the curious to the field of ML/AI. We explain the principles as commonly used methods as well as recent methodological advancements before we discuss risks and what we see as future directions of the field. Finally, we show practical examples of neuroscience to illustrate the use and limitations of ML., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier Inc. All rights reserved.)

Traditional Chinese medicine (TCM) has been recognized worldwide as a valuable asset of human medicine. The procedure of TCM is to treatment based on syndrome differentiation. However, the effect of TCM syndrome differentiation relies heavily on the experience of doctors. The gratifying progress of machine learning research in recent years has brought new ideas for TCM syndrome differentiation. In this paper, we propose a deep network model for TCM syndrome differentiation, which improves network performance by injecting TCM syndrome differentiation knowledge in the form of first-order logic into the deep network. Experimental results show that the accuracy of our proposed model reaches 89%, which is significantly better than the deep learning model MLP and other traditional machine learning models. In addition, we present the collected and formatted TCM syndrome differentiation (TSD) dataset, which contains more than 40,000 TCM clinical records. Moreover, 45 symptoms (""), 322 patterns(""), and more than 500 symptoms are labeled in TSD respectively. To the best of our knowledge, this is the first TCM syndrome differentiation dataset labeling diseases, syndromes and pattern. Such detailed labeling is helpful to explore the relationship between various elements of syndrome differentiation., Competing Interests: Declaration of competing interest We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript., (Copyright © 2023 Elsevier B.V. All rights reserved.)

Objective: Bio-Signals such as electroencephalography (EEG) and electromyography (EMG) are widely used for the rehabilitation of physically disabled people and for the characterization of cognitive impairments. Successful decoding of these bio-signals is however non-trivial because of the time-varying and non-stationary characteristics. Furthermore, existence of short- and long-range dependencies in these time-series signal makes the decoding even more challenging. State-of-the-art studies proposed Convolutional Neural Networks (CNNs) based architectures for the classification of these bio-signals, which are proven useful to learn spatial representations. However, CNNs because of the fixed size convolutional kernels and shared weights pay only uniform attention and are also suboptimal in learning short-long term dependencies, simultaneously, which could be pivotal in decoding EEG and EMG signals. Therefore, it is important to address these limitations of CNNs. To learn short- and long-range dependencies simultaneously and to pay more attention to more relevant part of the input signal, Transformer neural network-based architectures can play a significant role. Nonetheless, it requires a large corpus of training data. However, EEG and EMG decoding studies produce limited amount of the data. Therefore, using standalone transformers neural networks produce ordinary results. In this study, we ask a question whether we can fix the limitations of CNN and transformer neural networks and provide a robust and generalized model that can simultaneously learn spatial patterns, long-short term dependencies, pay variable amount of attention to time-varying non-stationary input signal with limited training data., Approach: In this work, we introduce a novel single hybrid model called ConTraNet, which is based on CNN and Transformer architectures that contains the strengths of both CNN and Transformer neural networks. ConTraNet uses a CNN block to introduce inductive bias in the model and learn local dependencies, whereas the Transformer block uses the self-attention mechanism to learn the short- and long-range or global dependencies in the signal and learn to pay different attention to different parts of the signals., Main Results: We evaluated and compared the ConTraNet with state-of-the-art methods on four publicly available datasets (BCI Competition IV dataset 2b, Physionet MI-EEG dataset, Mendeley sEMG dataset, Mendeley sEMG V1 dataset) which belong to EEG-HMI and EMG-HMI paradigms. ConTraNet outperformed its counterparts in all the different category tasks (2-class, 3-class, 4-class, 7-class, and 10-class decoding tasks)., Significance: With limited training data ConTraNet significantly improves classification performance on four publicly available datasets for 2, 3, 4, 7, and 10-classes compared to its counterparts., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Authors. Published by Elsevier Ltd.. All rights reserved.)

Grating-based X-ray phase contrast radiography and computed tomography (CT) are promising modalities for future medical applications. However, the ill-posed phase retrieval problem in X-ray phase contrast imaging has hindered its use for quantitative analysis in biomedical imaging. Deep learning has been proved as an effective tool for image retrieval. However, in practical grating-based X-ray phase contrast imaging system, acquiring the ground truth of phase to form image pairs is challenging, which poses a great obstacle for using deep leaning methods. Transfer learning is widely used to address the problem with knowledge inheritance from similar tasks. In the present research, we propose a virtual differential absorption model and generate a training dataset with differential absorption images and absorption images. The knowledge learned from the training is transferred to phase retrieval with transfer learning techniques. Numerical simulations and experiments both demonstrate its feasibility. Image quality of retrieved phase radiograph and phase CT slices is improved when compared with representative phase retrieval methods. We conclude that this method is helpful in both X-ray 2D and 3D imaging and may find its applications in X-ray phase contrast radiography and X-ray phase CT., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Background: The most commonly used therapy currently for inflammatory and autoimmune diseases is nonspecific anti-inflammatory drugs, which have various hazardous side effects. Recently, some anti-inflammatory peptides (AIPs) have been found to be a substitute therapy for inflammatory diseases like rheumatoid arthritis and Alzheimer's. Therefore, the identification of these AIPs is an emerging topic that is equally important., Methods: In this work, we have proposed an identification model for AIPs using a voting classifier. We used eight different feature descriptors and five conventional machine-learning classifiers. The eight feature encodings were concatenated to get a hybrid feature set. The five baseline models trained on the hybrid feature set were integrated via a voting classifier. Finally, a feature selection algorithm was used to select the optimal feature set for the construction of our final model, named IF-AIP., Results: We tested the proposed model on two independent datasets. On independent data 1, the IF-AIP model shows an improvement of 3%-5.6% in terms of accuracies and 6.7%-10.8% in terms of MCC compared to the existing methods. On the independent dataset 2, our model IF-AIP shows an overall improvement of 2.9%-5.7% in terms of accuracy and 8.3%-8.6% in terms of MCC score compared to the existing methods. A comparative performance analysis was conducted between the proposed model and existing methods using a set of 24 novel peptide sequences. Notably, the IF-AIP method exhibited exceptional accuracy, correctly identifying all 24 peptides as AIPs. The source code, pre-trained models, and all datasets are made available at https://github.com/Mir-Saima/IF-AIP., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023. Published by Elsevier Ltd.)

Identifying the most relevant variables or features in massive datasets for dimensionality reduction can lead to improved and more informative display, faster computation times, and more explainable models of complex systems. Despite significant advances and available algorithms, this task generally remains challenging, especially in unsupervised settings. In this work, we propose a method that constructs correlation networks using all intervening variables and then selects the most informative ones based on network bootstrapping. The method can be applied in both supervised and unsupervised scenarios. We demonstrate its functionality by applying Uniform Manifold Approximation and Projection for dimensionality reduction to several high-dimensional biological datasets, derived from 4D live imaging recordings of hundreds of morpho-kinetic variables, describing the dynamics of thousands of individual leukocytes at sites of prominent inflammation. We compare our method with other standard ones in the field, such as Principal Component Analysis and Elastic Net, showing that it outperforms them. The proposed method can be employed in a wide range of applications, encompassing data analysis and machine learning., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Generative models, such as Generative Adversarial Networks (GANs), have recently shown remarkable capabilities in various generation tasks. However, the success of these models heavily depends on the availability of a large-scale training dataset. When the size of the training dataset is limited, the quality and diversity of the generated results suffer from severe degradation. In this paper, we propose a novel approach, Reverse Contrastive Learning (RCL), to address the problem of high-quality and diverse image generation under few-shot settings. The success of RCL benefits from a two-sided, powerful regularization. Our proposed regularization is designed based on the correlation between generated samples, which can effectively utilize the latent feature information between different levels of samples. It does not require any auxiliary information or augmentation techniques. A series of qualitative and quantitative results show that our proposed method is superior to the existing State-Of-The-Art (SOTA) methods under the few-shot setting and is still competitive under the low-shot setting, showcasing the effectiveness of RCL. Code will be released upon acceptance at https://github.com/gouayao/RCL., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Capturing global and subtle discriminative information using attention mechanisms is essential to address the challenge of inter-class high similarity for vehicle re-identification (Re-ID) task. Mixing self-information of nodes or modeling context based on pairwise dependencies between nodes are the core ideas of current advanced attention mechanisms. This paper aims to explore how to utilize both dependency context and self-context in an efficient way to facilitate attention to learn more effectively. We propose a heterogeneous context interaction (HCI) attention mechanism that infers the weights of nodes from the interactions of global dependency contexts and local self-contexts to enhance the effect of attention learning. To reduce computational complexity, global dependency contexts are modeled by aggregating number-compressed pairwise dependencies, and the interactions of heterogeneous contexts are restricted to a certain range. Based on this mechanism, we propose a heterogeneous context interaction network (HCI-Net), which uses channel heterogeneous context interaction module (CHCI) and spatial heterogeneous context interaction module (SHCI), and introduces a rigid partitioning strategy to extract important global and fine-grained features. In addition, we design a non-similarity constraint (NSC) that forces the HCI-Net to learn diverse subtle discriminative information. The experiment results on two large datasets, VeRi-776 and VehicleID, show that our proposed HCI-Net achieves the state-of-the-art performance. In particular, the mean average precision (mAP) reaches 83.8% on VeRi-776 dataset., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Large deep learning models are impressive, but they struggle when real-time data is not available. Few-shot class-incremental learning (FSCIL) poses a significant challenge for deep neural networks to learn new tasks from just a few labeled samples without forgetting the previously learned ones. This setup can easily leads to catastrophic forgetting and overfitting problems, severely affecting model performance. Studying FSCIL helps overcome deep learning model limitations on data volume and acquisition time, while improving practicality and adaptability of machine learning models. This paper provides a comprehensive survey on FSCIL. Unlike previous surveys, we aim to synthesize few-shot learning and incremental learning, focusing on introducing FSCIL from two perspectives, while reviewing over 30 theoretical research studies and more than 20 applied research studies. From the theoretical perspective, we provide a novel categorization approach that divides the field into five subcategories, including traditional machine learning methods, meta learning-based methods, feature and feature space-based methods, replay-based methods, and dynamic network structure-based methods. We also evaluate the performance of recent theoretical research on benchmark datasets of FSCIL. From the application perspective, FSCIL has achieved impressive achievements in various fields of computer vision such as image classification, object detection, and image segmentation, as well as in natural language processing and graph. We summarize the important applications. Finally, we point out potential future research directions, including applications, problem setups, and theory development. Overall, this paper offers a comprehensive analysis of the latest advances in FSCIL from a methodological, performance, and application perspective., Competing Interests: Declaration of competing interest The authors declare no conflict of interests., (Copyright © 2023 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Graph Convolutional Network (GCN) has become a hotspot in graph-based machine learning due to its powerful graph processing capability. Most of the existing GCN-based approaches are designed for single-view data. In numerous practical scenarios, data is expressed through multiple views, rather than a single view. The ability of GCN to model homogeneous graphs is indisputable, while it is insufficient in facing the heterophily property of multi-view data. In this paper, we revisit multi-view learning to propose an implicit heterogeneous graph convolutional network that efficiently captures the heterogeneity of multi-view data while exploiting the powerful feature aggregation capability of GCN. We automatically assign optimal importance to each view when constructing the meta-path graph. High-order cross-view meta-paths are explored based on the obtained graph, and a series of graph matrices are generated. Combining graph matrices with learnable global feature representation to obtain heterogeneous graph embeddings at various levels. Finally, in order to effectively utilize both local and global information, we introduce a graph-level attention mechanism at the meta-path level that allocates private information to each node individually. Extensive experimental results convincingly support the superior performance of the proposed method compared to other state-of-the-art approaches., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

The Golgi body is a critical organelle in eukaryotic cells responsible for processing and modifying proteins and lipids. Under certain conditions, such as stress, disease, or ageing, the Golgi structure alters. Therefore, understanding the mechanisms that regulate Golgi dispersion has significant research contributions to identifying disease. However, there is a lack of tools to quantify the Golgi dispersion datasets. In this paper, we aim to automate the process of quantification of Golgi dispersion and use extracted features to classify dispersed Golgi images from undispersed Golgi images using machine learning models. First, we collected confocal microscopy images of transiently transfected HeLa cells expressing Galactose-1-phosphate uridylyltransferase (GALT)- green fluorescent protein (GFP) to quantify Golgi dispersal and classification. For the quantification, we introduced automated image processing and segmentation by applying mean and Gaussian filters. Then we used Otsu thresholding on preprocessed images and watershed segmentation to refine the segmentation of dispersed Golgi particles. In the case of classification, we extracted features from the Golgi dispersal images and classified them into empty vector (EV) versus CARP1 ring mutant (CARP1 RM) and empty vector (EV) versus CARP1 wildtype (CARP1 WT) classes. Our approach used machine-learning models, including logistic regression, decision tree, random forest, Naive Bayes, k-Nearest Neighbor (KNN), and gradient boosting for dispersed Golgi image classification. The experiment results show that our quantification technique on Golgi dispersal images reached 65% classification accuracy when the system uses a gradient boosting classifier for EV vs. CARP1 WT classification. Furthermore, our approach achieved 65% classification accuracy using a random forest classifier for EV vs. CARP1 RM classification., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Accurately modelling crashes, and predicting crash occurrence and associated severities are a prerequisite for devising countermeasures and developing effective road safety management strategies. To this end, crash prediction modelling using machine learning has evolved over two decades. With the advent of big data that provides unprecedented opportunities to better understand the crash mechanism and its determinants, such efforts will likely be accelerated. To gear these efforts, understanding state-of-the-art machine learning-based crash prediction models becomes paramount to summarise the lessons learned from past efforts, which can assist in developing robust and accurate models. This review paper aims to address this gap by systematically reviewing the machine learning studies on crash modelling. Models are reviewed from three aspects of the application: (a) crash occurrence (or real-time crash) prediction, (b) crash frequency prediction, and (c) injury severity prediction. Further, model intricacies that impact model performance are identified and thoroughly reviewed. This comprehensive review highlights specific gaps and future research needs in three aforementioned model applications, such as improper selection of non-crash events for crash occurrence models, the inability of future forecasting of crash frequency models, and inconsistency in injury severity classes. Critical research needs relating to model development, evaluation, and application are also discussed. This review envisages methodological advancements in machine learning models for crash prediction modelling and leveraging big data to better link crashes with its determinants., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Aptamers are increasingly recognized as potent alternatives to antibodies for diagnostic and therapeutic applications. The application of deep learning, particularly attention-based models, for aptamer (DNA/RNA) sequences is an innovative field. The ongoing advancements in aptamer sequencing technologies coupled with machine learning algorithms have resulted in novel developments. Further research is required to investigate the full potential of deep learning models and address the challenges associated with the generation of sequences, like the large search space of possible sequences. In this study, we propose a workflow that integrates an attention mechanism within a framework of a generative variational autoencoder, to generate novel sequences by expanding latent memory. They show 100 % novelty compared with the dataset, and approximately 88 % of them show negative values for the minimum free energy, which may indicate the likelihood of an RNA sequence folding into a functional structure. Because the field of aptamer discovery is affected by data scarcity, advanced strategies that facilitate the generation of diverse and superior sequences are necessitated. The utilization of our workflow can result in novel aptamers. Thus, investigations such as the present study can address the abovementioned challenge. Our research is anticipated to facilitate further discoveries and advancements in aptamer fields., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 Elsevier B.V. All rights reserved.)

Gene expression is regulated by transcription factors that work together to read cis-regulatory DNA sequences. The 'cis-regulatory code' - how cells interpret DNA sequences to determine when, where and how much genes should be expressed - has proven to be exceedingly complex. Recently, advances in the scale and resolution of functional genomics assays and machine learning have enabled substantial progress towards deciphering this code. However, the cis-regulatory code will probably never be solved if models are trained only on genomic sequences; regions of homology can easily lead to overestimation of predictive performance, and our genome is too short and has insufficient sequence diversity to learn all relevant parameters. Fortunately, randomly synthesized DNA sequences enable testing a far larger sequence space than exists in our genomes, and designed DNA sequences enable targeted queries to maximally improve the models. As the same biochemical principles are used to interpret DNA regardless of its source, models trained on these synthetic data can predict genomic activity, often better than genome-trained models. Here we provide an outlook on the field, and propose a roadmap towards solving the cis-regulatory code by a combination of machine learning and massively parallel assays using synthetic DNA., (© 2023. Springer Nature Limited.)

• An improved safety potential field, named the Work-Zone Crash Risk Field (WCRF), is developed for real-time crash risk assessment in highway work zones. • The WCRF performs better than conventional safety surrogate measures in identifying high-risk vehicles in work zones. • The impact of setting up a work zone on the crash risk of each lane at transition areas is analyzed. • Several key traffic features that may contribute to crash risk in transition areas of the work zones are identified. • The heterogeneity in the impact of risk-contributing traffic features over different work zones is discussed. Due to complex traffic conditions, transition areas in highway work zones are associated with a higher crash risk than other highway areas. Understanding risk-contributing features in transition areas is essential for ensuring traffic safety on highways. However, conventional surrogate safety measures (SSMs) are quite limited in identifying the crash risk in transition areas due to the complex traffic environment. To this end, this study proposes an improved safety potential field, named the Work-Zone Crash Risk Field (WCRF). The WCRF force can be used to measure the crash risk of individual vehicles that enter a work zone considering the influence of multiple features, upon which the overall crash risk of the road segment in a specific time window can be estimated. With the overall crash risk used as a label, the time-window-based traffic data are used to train and validate an eXtreme Gradient Boosting (XGBoost) classifier, and the Shapley Additive Explanations (SHAP) method is integrated with the XGBoost classifier to identify the key risk-contributing traffic features. To assess the proposed approach, a case study is conducted using real-time vehicle trajectory data collected in two work zones along a highway in China. The results demonstrate that the WCRF-based SSM outperforms conventional SSMs in identifying crash risks in work zone transition areas on highways. In addition, we perform lane-based analysis regarding the impact of setting up work zones on highway safety and investigate the heterogeneity in risk-contributing features across different work zones. Several interesting findings from the analysis are reported in this paper. Compared to existing SSMs, the WCRF-based SSM offers a more practical and comprehensive way to describe the crash risk in work zones. The approach using the developed WCRF technique offers improved capabilities in identifying key risk-contributing features, which is expected to facilitate the development of safety management strategies for work zones. [ABSTRACT FROM AUTHOR]

• We apply Latent Dirichlet Allocation algorithm to discover activity patterns from floating vehicle trajectory data. • We incorporate activity patterns into traffic safety analysis to supplement traditional traffic exposure measures. • We use Graph Convolutional Network to explore multiple factors influencing spatial patterns of traffic crashes. • We estimate non-linear relationships between variables and traffic crashes using interpretive methodologies. In recent years, increased attention has been given to understanding the spatial pattern of crashes in urban areas. Accurately capturing the spatial relationship between crash counts and variables requires extracting hidden information from multiple data sources. In this study, we propose a machine learning model to explore the spatial impact of activity patterns on spatially aggregated crash counts. Our paper introduces a two-step framework: (a) the Latent Dirichlet Allocation (LDA) model, an unsupervised method for mining hidden activity patterns from floating vehicle trajectory data, and (b) the Graph Convolutional Network (GCN) model, which builds the spatial relationship between multi-source data. The data and hidden activity patterns were aggregated into 175 Traffic Analysis Zones (TAZs) in San Francisco using spatial partitioning. The GCN model provided higher prediction accuracy than commonly used machine learning algorithms that did not consider combined spatial relationships and those that only considered traditional vehicle counts data. Furthermore, we used attribution algorithms to obtain the respective weight scores of each factor. Our results reveal that daily vehicle kilometers traveled, road density, population density, commercial activity during weekends, and residential activity during morning peak hours on weekdays are factors associated with crashes. [ABSTRACT FROM AUTHOR]

Spiking Neural Networks (SNNs) are energy-efficient artificial neural network models that can carry out data-intensive applications. Energy consumption, latency, and memory bottleneck are some of the major issues that arise in machine learning applications due to their data-demanding nature. Memristor-enabled Computing-In-Memory (CIM) architectures have been able to tackle the memory wall issue, eliminating the energy and time-consuming movement of data. In this work we develop a scalable CIM-based SNN architecture with our fabricated two-layer memristor crossbar array. In addition to having an enhanced heat dissipation capability, our memristor exhibits substantial enhancement of 10% to 66% in design area, power and latency compared to state-of-the-art memristors. This design incorporates an inter-spike interval (ISI) encoding scheme due to its high information density to convert the incoming input signals into spikes. Furthermore, we include a time-to-first-spike (TTFS) based output processing stage for its energy-efficiency to carry out the final classification. With the combination of ISI, CIM and TTFS, this network has a competitive inference speed of 2μs/image and can successfully classify handwritten digits with 2.9mW of power and 2.51pJ energy per spike. The proposed architecture with the ISI encoding scheme can achieve ∼10% higher accuracy than those of other encoding schemes in the MNIST dataset. [ABSTRACT FROM AUTHOR]

The wireless sensor network (WSN) has been gaining popularity for automation and performance improvement in different IoT-based applications. The resource-constrained nature and operating environment of IoT make the devices highly vulnerable to different attacks. However, the Physically Unclonable Function (PUF) helps to implement secure and lightweight authentication protocols for IoT. In this context, few computation-intensive authentication protocols are found in the literature that have addressed secure IoT communication in WSN. Besides, these protocols depend on the local storage of PUF-CRP, which is susceptible to security attacks. This work proposes a lightweight and secure authentication protocol for the IoT devices in WSN. A PUF and its machine learning (ML)–based soft model is integrated to ensure secure authentication and lightweight computation in WSN. PUF prevents physical attacks while carrying much less hardware fingerprints, and the ML-based PUF provides the desired resiliency against PUF identity-based attacks by eliminating the requirement of CRP-based storage. The proposed mechanism delivers two-way authentication while nullifying the attacks on IoT. The proposed protocol is implemented on Xilinx Artix-7 FPGA and Raspberry Pi for testability and performance evaluation. Experiment results and analysis signify its low-cost computations and lightweight features desired for IoT. [ABSTRACT FROM AUTHOR]

Cyber-Physical Systems (CPS) continuously interact with their physical environments through embedded software controllers that observe the environments and determine actions. Field Operational Tests (FOT) are essential to verify to what extent the CPS under analysis can achieve certain CPS goals, such as satisfying the safety and performance requirements, while interacting with the real operational environment. However, performing many FOTs to obtain statistically significant verification results is challenging due to its high cost and risk in practice. Simulation-based verification can be an alternative to address the challenge, but it still requires an accurate virtual environment model that can replace the real environment interacting with the CPS in a closed loop. In this article, we propose ENVI (ENVironment Imitation), a novel approach to automatically generate an accurate virtual environment model, enabling efficient and accurate simulation-based CPS goal verification in practice.To do this, we first formally define the problem of the virtual environment model generation and solve it by leveraging Imitation Learning (IL), which has been actively studied in machine learning to learn complex behaviors from expert demonstrations. The key idea behind the model generation is to leverage IL for training a model that imitates the interactions between the CPS controller and its real environment as recorded in (possibly very small) FOT logs. We then statistically verify the goal achievement of the CPS by simulating it with the generated model. We empirically evaluate ENVI by applying it to the verification of two popular autonomous driving assistant systems. The results show that ENVI can reduce the cost of CPS goal verification while maintaining its accuracy by generating accurate environment models from only a few FOT logs. The use of IL in virtual environment model generation opens new research directions, further discussed at the end of the article. [ABSTRACT FROM AUTHOR]

Graph Neural Networks (GNNs) are a popular machine learning framework for solving various graph processing applications. This framework exploits both the graph topology and the feature vectors of the nodes. One of the important applications of GNN is in the semi-supervised node classification task. The accuracy of the node classification using GNN depends on (i) the number and (ii) the choice of the training nodes. In this article, we demonstrate that increasing the training nodes by selecting nodes from the same class that are spread out across non-contiguous subgraphs, can significantly improve the accuracy. We accomplish this by presenting a novel input intervention technique that can be used in conjunction with different GNN classification methods to increase the non-contiguous training nodes and, thereby, improve the accuracy. We also present an output intervention technique to identify misclassified nodes and relabel them with their potentially correct labels. We demonstrate on real-world networks that our proposed methods, both individually and collectively, significantly improve the accuracy in comparison to the baseline GNN algorithms. Both our methods are agnostic. Apart from the initial set of training nodes generated by the baseline GNN methods, our techniques do not need any other extra knowledge about the classes of the nodes. Thus, our methods are modular and can be used as pre-and post-processing steps with many of the currently available GNN methods to improve their accuracy. [ABSTRACT FROM AUTHOR]

Emergency and crisis simulations play a pivotal role in equipping authorities worldwide with the necessary tools to minimize the impact of catastrophic events. Various studies have explored the integration of intelligence into Multi-Agent Systems (MAS) for crisis simulation. This involves incorporating psychological behaviours from the social sciences and utilizing data-driven machine learning models with predictive capabilities. A recent advancement in behavioural modelling is the Conscious Movement Model (CMM), designed to modulate an agent's movement patterns dynamically as the situation unfolds. Complementing this, the model incorporates a Conscious Movement Memory-Attention (CMMA) mechanism, enabling learnability through training on pedestrian trajectories extracted from video data. The CMMA facilitates mapping a pedestrian's attention to their surroundings and understanding how their past decisions influence their subsequent actions. This study proposes an efficient framework that integrates the trained CMM into a simulation model specifically tailored for emergency evacuations, ensuring realistic outcomes. The resulting simulation framework automates strategy management and planning for diverse emergency evacuation scenarios. A single-objective method is presented for generating prescriptive analytics, offering effective strategy options based on predefined operational rules. To validate the framework's efficacy, a case study of a theatre evacuation is conducted. In essence, this research establishes a robust simulation framework for crisis management, with a particular emphasis on modelling pedestrians during emergency evacuations. The framework generates prescriptive analytics to aid authorities in executing rescue and evacuation operations effectively. [ABSTRACT FROM AUTHOR]

Recently, machine learning-based image compression has attracted increasing interests and is approaching the state-of-the-art compression ratio. But unlike traditional codec, it lacks a universal optimization method to seek efficient representation for different images. In this paper, we develop a plug-and-play optimization framework for seeking higher compression ratio, which can be flexibly applied to existing and potential future compression networks. To make the latent representation more efficient, we propose a novel latent optimization algorithm to adaptively remove the redundancy for each image. Additionally, inspired by the potential of side information for traditional codecs, we introduce side information into our framework, and integrate side information optimization with latent optimization to further enhance the compression ratio. In particular, with the joint side information and latent optimization, we can achieve fine rate control using only single model instead of training different models for different rate-distortion trade-offs, which significantly reduces the training and storage cost to support multiple bit rates. Experimental results demonstrate that our proposed framework can remarkably boost the machine learning-based compression ratio, achieving more than 10% additional bit rate saving on three different representative network structures. With the proposed optimization framework, we can achieve 7.6% bit rate saving against the latest traditional coding standard VVC on Kodak dataset, yielding the state-of-the-art compression ratio. [ABSTRACT FROM AUTHOR]

Enabled by the increasingly available embedded hardware accelerators, the capability of executing advanced machine learning models at the edge of the Internet of Things (IoT) triggers interest of applying Artificial Intelligence of Things (AIoT) systems for industrial applications. The in situ inference and decision made based on the sensor data allow the industrial system to address a variety of heterogeneous, local-area non-trivial problems in the last hop of the IoT networks. Such a scheme avoids the wireless bandwidth bottleneck and unreliability issues, as well as the cumbersome cloud. However, the literature still lacks presentations of industrial AIoT system developments that provide insights into the challenges and offer lessons for the relevant research and industry communities. In light of this, we present the design, deployment, and evaluation of an industrial AIoT system for improving the quality control of HP Inc.'s ink cartridge manufacturing lines. While our development has obtained promising results, we also discuss the lessons learned from the whole course of the work, which could be useful to the development of other industrial AIoT systems for quality control in manufacturing. [ABSTRACT FROM AUTHOR]