Summary COVID‐19 is a novel coronavirus‐induced disease and automatic identification of COVID‐19 using computer‐assisted methods can facilitate faster diagnostic efficiency. Current research typically employs a single model for COVID‐19 identification, while implicit and complementary knowledge between heterogeneous networks is neglected. To address these issues, we propose a new model based on deep mutual learning with online feature alignment called DML‐OFA to more effectively diagnose COVID‐19. First, we use a traditional deep mutual learning (DML) framework to allow two parallel heterogeneous networks to learn from each other to form two effective feature extractors. In addition, we embed the adaptive feature fusion classifier and logits ensembling module in the proposed DML‐OFA, which can simultaneously learn implicit complementary knowledge from feature maps and logits. We evaluated DML‐OFA on four public datasets: Covid‐chestxray‐dataset, ChestXRay2017, Coronavirus‐dataset and COVIDx. The results showed that our model attains 97.10 Accuracy, 97.28 Specificity, 96.21 Recall, 97.45 Precision, and 96.82 F1‐score, which outperforms other previous related works.

Pneumonia is one of the largest causes of death in the world. Deep learning techniques can assist doctors to detect the areas of pneumonia in the chest X-rays images. However, existing methods lack sufficient consideration for the large variation scale and the blurred boundary of the pneumonia area. Here, we present a deep learning method based on Retinanet for pneumonia detection. First, we introduce Res2Net into Retinanet to get the multi-scale feature of pneumonia. Then, we proposed a novel predicted boxes fusion algorithm, named Fuzzy Non-Maximum Suppression (FNMS), which gets a more robust predicted box by fusing the overlapping detection boxes. Finally, we get the performance outperforms than existing methods by integrating two models with different backbones. We report the experimental result in the single model case and the model ensemble case. In the single model case, RetinaNet with FNMS algorithm and Res2Net backbone is better than RetinaNet and other models. In the model ensemble case, the final score of predicted boxes that fused by the FNMS algorithm is better than NMS, Soft-NMS, and weighted boxes fusion. Experimental results on the pneumonia detection dataset verify the superiority of the FNMS algorithm and the proposed method in the pneumonia detection task.

Training data are often imbalanced across classes in practice, and such class imbalance issue often causes model predictions biased toward majority classes during inference. Different from existing solutions which employ various training strategies to alleviate the class imbalance issue, this study proposes a novel two-head model architecture to help alleviate the issue. One auxiliary classifier head helps the feature extractor of the classifier more fairly learn to extract features for each class, and the main classifier head learns in a more class-balanced manner by dividing each majority class into multiple clusters in advance and considering each cluster as a new class. Extensive empirical evaluations on four class-imbalanced image datasets showed that the proposed approach achieves state-of-the-art classification performance.

Graph data can characterize many complex network systems, hence many community detection methods have been proposed to analyze graph problems. Recently, Non-negative matrix factorization (NMF) has been widely used for community detection due to its good interpretability of community detection results. However, existing NMF-based community detection methods are designed without considering extreme adversarial attacks (or perturbations). In reality, complex and diverse adversarial perturbations can easily destroy the network structure, so the classical NMF method cannot effectively identify the attacked network. Inspired by adversarial training technique, we propose a novel NMF method for extreme adversarial attacks, coined as Extreme Adversarial Attack NMF. Specifically, we consider the existence of extreme adversarial perturbations which destroy the network structure, while the matrix factorization results can still fit the attacked network well. A strict optimization algorithm is also devised to dynamically simulate the generation of extreme adversarial perturbations. This work aims to improve the robustness and generalization of NMF methods in graph-based learning. We conduct extensive experiments on six original real-world networks and their edge-missing networks. The experiment results show that the proposed method can effectively improve the performance of the NMF-based community detection method, which outperforms 4.69% and 7.07% on average over the standard NMF on the original and edge-missing networks, respectively.

In mobile crowdsensing activities, it is usually necessary for participants to upload sensing data and related locations. The existing location privacy-preserving mechanisms cannot well protect a user's trajectory privacy because attackers can mine the user's trajectory features through data analysis techniques. Aiming at the trajectory privacy protection problem, this article proposes a differential location privacy-preserving mechanism based on trajectory obfuscation (LPMT). LPMT first extracts the stay points as the features of a trajectory based on the sliding window algorithm, and then obfuscates each stay point to a target obfuscation subregion through the exponential mechanism, and finally performs the Laplace sampling in the target obfuscation subregion to obtain the obfuscated GPS points. Compared with the baseline mechanisms, LPMT can reduce data quality loss by more than 20% while providing the same level of obfuscation quality, which indicates that LPMT has the advantages of strong security and high quality of service.

The diversity of base classifiers and integration of multiple classifiers are two key issues in the field of ensemble learning. This paper puts forward a hybrid ensemble algorithm combining AdaBoost and genetic algorithm(GA) for cancer classification with gene expression data. The decision group is designed to increase the diversity of base classifier pool, and the GA is used to assign weight to each base classifier, thus to improve the classification performance by avoiding local extrema. The decision groups composed by using base classifiers, including K-nearest neighbor (KNN), Naïve Bayes (NB), and Decision Tree (C4.5). Experimental results show that the proposed algorithm is superior to those existing ensemble learning methods, such as Bagging, Random Forest (RF), Rotation Forest (RoF), AdaBoost, AdaBoost-BPNN, AdaBoost-SVM, and AdaBoost-RF, especially it has better performance on small samples and unbalanced gene expression data processing.

In real-world applications of hyperspectral images (HSIs), we may frequently face the following situation: two HSI scenes (named source and target scenes, respectively) contain similar land over objects, but they are captured in different spots or at different time. Even if they are captured by the same hyper-spectral sensor, there exist spectral shift between them. In our previous work, we tried to solve the spectral shift by dictionary sharing or domain-invariant feature selection. However, a more regular case is that two similar HSI scenes are captured by different hyperspectral sensors. How to mine the relationship between such HSIs is a more challenging problem, since the feature spaces are totally different. A natural approach is to learn a unified low-dimensional feature subspace which can bridge the two HSI scenes. In this work, we propose a dual dictionary nonnegative matrix factorization (DDNMF) algorithm for the aforementioned goal. In details, an individual domain-specific dictionary is learned for each scene, and two dictionary learning tasks (for source and target scenes) are coupled by manifold regularization, ensuring that pixels belonging to a same land cover class have similar representations over the learned dictionaries, even if they come from different scenes. Experimental results show that the proposed algorithm can indeed mine a unified feature subspace shared between two different HSI scenes.

Feature extraction is an important research topic in hyper-spectral image (HSI) classification. However, most of feature extraction methods only extract low-level features, which makes them not perform well in the applications of HSI. In this paper, we have proposed a non-negative matrix factorization (NMF) based deep feature extraction algorithm, namely deep NMF. Deep NMF tries to construct a deep feature representation by cascading multiple NMFs. Reconstruction residual of NMF is passed layer by layer to reduce information loss. Meanwhile, passing residuals between layers can construct a feature hierarchy from coarse to fine. Furthermore, activation functions are applied between adjacent layers to enhance the ability of non-linear feature extraction. Experimental results have also shown that our algorithm is computationally efficient and effective for HSI classification.

BACKGROUND: Cost-sensitive algorithm is an effective strategy to solve imbalanced classification problem. However, the misclassification costs are usually determined empirically based on user expertise, which leads to unstable performance of cost-sensitive classification. Therefore, an efficient and accurate method is needed to calculate the optimal cost weights. RESULTS: In this paper, two approaches are proposed to search for the optimal cost weights, targeting at the highest weighted classification accuracy (WCA). One is the optimal cost weights grid searching and the other is the function fitting. Comparisons are made between these between the two algorithms above. In experiments, we classify imbalanced gene expression data using extreme learning machine to test the cost weights obtained by the two approaches. CONCLUSIONS: Comprehensive experimental results show that the function fitting method is generally more efficient, which can well find the optimal cost weights with acceptable WCA.

The extreme learning machine (ELM) is famous for its single hidden-layer feed-forward neural network which results in much faster learning speed comparing with traditional machine learning techniques. Moreover, extensions of ELM achieve stable classification performances for imbalanced data. In this paper, we introduce a hybrid method combining the extended Kalman filter (EKF) with cost-sensitive dissimilar ELM (CS-D-ELM). The raw data are preprocessed by EKF to produce inputs for the CS-D-ELM classifier. Experimental results show that the proposed method is more suitable for real-time fault diagnosis of air handling units than traditional approaches.

Feature selection is an important research topic for hyperspectral images (HSIs). It helps to remove the noisy or redundant features. Traditional feature selection algorithms are mostly performed within a single HSI scene (dataset). However, appearance of massive HSIs requires the feature selection problems to be considered across different HSI scenes, e.g., two HSI scenes obtained from different spots or at different time. In this case, the features are not identically distributed within two scenes due to spectral shift. To solve this problem, a cross-scene feature selection algorithm is proposed in this work for HSIs, which is based on cross-domain information gain (CDIG). The main motivation includes two factors, one is the discriminant of selected features to separate different land-cover classes, while the other is the consistency of the selected features between different scenes. Consequently, the proposed CDIG reaches a compromise between aforementioned two factors. Experimental results on two cross-scene HSI datasets show the advantages of the proposed CDIG in cross-scene feature selection problems.

Summary form only given. In HEVC video coding, all-zero-block (AZB) detection is an efficient tool to decrease the complexity of mode decision for rate distortion optimization (RDO). Threshold based on deadzone quantization was widely used for AZB detection, and the threshold was derived according to ensemble based statistical analysis however using individual sample's parameters such as block's SATD [1] [2]. This paper proposes an adaptive AZB algorithm targeting for RDO quantization (RDOQ) as shown in Fig.1. Inspired by Bayesian decision, this paper proposes an more accurate zero-quantized deadzone offset model (ΓDCT) shown in Fig.2, which is formulated as function of quantization parameter Qp and DCT coefficient distribution parameter Λ. Then, a local parameter depicting the individual block's characteristics regarding the inter-coefficient distribution is combined jointly with SATD to derive an adaptive AZB detection threshold model, which is implemented by comparing with adaptive threshold ξσF instead of conventionally fixed threefold standard deviation σF. The experimental results demonstrate that the proposed work detects 90.8% AZB with smaller than 2.66% false alarm rate on average, with negligible rate distortion performance loss. This work is well-suited for fast RD optimized HEVC coding.

In high definition (HD) video coders, huge memory access bandwidth is the major throughput bottleneck. Lossless embedded compression is an efficient solution to alleviate the bandwidth burden, in which image are compressed before writing into local memory and decompressed after retrieving from local memory. This study proposes a hardware‐oriented lossless image compression algorithm, supporting block and line random access flexibly for adapting diverse hardware video codec architectures. The major contributions are characterised as follows. First, block or pixel‐level adaptive prediction is proposed to fully utilise the image spatial correlation by employing adaptive mode decision. Second, multiple‐range semi‐fixed (SF) variable length coding (VLC) is employed to describe the prediction residue, and adaptive block size selection is employed for SF VLC to fully utilise the statistical redundancy. In addition, Huffman VLC is further employed to represent the control syntax elements. Third, four‐stage pipeline hardware architecture is proposed to implement the proposed algorithm. Simulation results show that the proposed algorithm achieves competitive rate compression performance compared with reference algorithms. The proposed hardware architecture is verified supporting real‐time processing for quad‐HD videos at the frequency of 166 MHz. The proposed work achieves reducing memory access bandwidth by ∼55.2%, which is useful for hardwired video coding.

Aiming at the increasing security risks of children, this paper presents and implements a kind of Mobile Children Security Monitoring (MCSM) system based on android phones to help guardian to acquire whether children are safe or not. MCSM implements the software hand function and the danger zone function for two typical safety scenarios, i.e., going outside with their guardians and without their guardians respectively. The software hand function can keep children in guardian's view by using Bluetooth near field communication, and the safety zone function can make guardians know children's location timely by using GPS sensors, acceleration sensors, and mobile GIS (Geographic Information System). Experiments shows the system has the characteristics of high reliability, short response time and high accuracy, and can meet the requirements to ensure children's safety.

How to manage cross-enterprise processes in a flexible and efficient way has grown to be an urgent problem needed to be solved in BPM (business process management). Workflow technology is the most popular approach to enable the automation of business processes. However, traditional workflow management systems do no manage cross-enterprise processes well. This paper analyzes the characteristics of cross-enterprise processes and points out challenges for traditional workflow systems. It proposes a service-oriented workflow management system framework, which makes use of the web service technology to deal with those distributed and heterogeneous applications. This framework provides an applicable mechanism to manage cross-enterprise processes in an efficient way.

There is great potential in mining gene expression microarray databases to discover causal relationships in the gene-regulation pathway. Several methods using Bayesian networks have been reported. Most of them use a heuristics search based on the criteria for choosing a network, but these methods are often computationally intractable for microarray data with thousands of genes. In this work, a simple constrained-based, local causal discovery method is presented. This method is computationally feasible but does not attempt to discover complete causal structure. To show the effectiveness of this method, we have conducted simulations and applied this method to the data set from Hughes et al. from 300 expression profiles of yeast. Using this method, results of simulation data tests demonstrated that the accuracy ratios of causal relationships became higher when the sample size increased. From the yeast data set, a number of causal relations were found. A cursory analysis shows some of the relations have biological sense, others need further investigation.