Dielectric elastomer-driven tunable prisms are novel liquid optical devices that enable two degree-of-freedom beam deflection. However, their periodic dynamic response exhibits hysteresis, creep, and rate-dependent viscoelastic behavior, which makes modeling the nonlinear dynamics and precise control challenging. This paper analyzes the dynamics and kinematics of a biconical tunable prism. Based on the principle of non-equilibrium thermodynamics, a nonlinear dynamic model of the tunable prism is established, taking into account the geometric configuration of the prism, the viscoelasticity of the DE, and the liquid resistance. We developed feedforward controller based on the model and combined with a PI feedback controller to form the hybrid controller. Experimental results show that the theoretical model can accurately predict the complex dynamic response of the system, with root mean square error of prediction lower than 5.59%. Single and dual axis deflection angle tracking root mean square error does not exceed 1.25%. These results demonstrate the effectiveness of the proposed dynamic modeling and tracking control method.

Multi-view clustering has exhibited exceptional performance by harnessing complementary information and extracting shared semantics across diverse views. However, existing methods still encounter substantial obstacles in exploiting high-quality consensus representation: (i) they typically lack effective guidance mechanisms under unsupervised conditions, leading to suboptimal representation learning, and (ii) they overlook the intrinsic cluster structure when forming sample pairs to align representation across views, conflicting with clustering objectives. To address these issues, we propose a contrastive-based method calledPseudo-LabelSimilarityGraph-driven Multi-ViewClustering (PSGVC). Specifically, to uncover the potential semantic relationships between samples, we construct a pseudo-label similarity graph based on the consensus representation. Then, we design a similarity graph-based contrastive loss, which utilizes the pseudo-label similarity graph as a global supervision signal to guide the optimization of cross-view embedded representation, thereby indirectly improving the quality of consensus representation. Additionally, we propose a weighted cluster-aware contrastive loss to align the consensus representation with view-specific representation. It leverages the cluster structure contained in the pseudo-label similarity graph to achieve cluster-level contrastive learning, further enhancing the quality of the consensus representation. Extensive experimental results show that our PSGVC achieves state-of-the-art clustering performance across multiple datasets.

Multimodal unsupervised domain adaptation leverages un-labeled data in the target domain to enhance multimodal systems continuously. While current state-of-the-art methods encourage interaction between sub-models of different modalities through pseudo-labeling and feature-level exchange, varying sample quality across modalities can lead to the propagation of inaccurate information, resulting in error accumulation. To address this, we propose Modal-Affinity Multimodal Domain Adaptation (MODfinity), a method that dynamically manages multimodal information flow through fine-grained control over teacher model selection, guiding information intertwining at both feature and label levels. By treating labels as an independent modality, MODfinity enables balanced performance assessment across modalities, employing a novel modal-affinity measurement to evaluate information quality. Additionally, we introduce a modal-affinity distillation technique to control sample-level information exchange, ensuring reliable multimodal interaction based on affinity evaluations within the feature space. Extensive experiments on three multimodal datasets demonstrate that our framework consistently outperforms state-of-the-art methods, particularly in high-noise environments.

Irregular multivariate time series (IMTS) are characterized by irregular time intervals within variables and unaligned observations across variables, posing challenges in learning temporal and variable dependencies. Many existing IMTS models either require padded samples to learn separately from temporal and variable dimensions, or represent original samples via bipartite graphs or sets. However, the former approaches often need to handle extra padding values affecting efficiency and disrupting original sampling patterns, while the latter ones have limitations in capturing dependencies among unaligned observations. To represent and learn both dependencies from original observations in a unified form, we propose HyperIMTS, a Hypergraph neural network for Irregular Multivariate Time Series forecasting. Observed values are converted as nodes in the hypergraph, interconnected by temporal and variable hyperedges to enable message passing among all observations. Through irregularity-aware message passing, HyperIMTS captures variable dependencies in a time-adaptive way to achieve accurate forecasting. Experiments demonstrate HyperIMTS’s competitive performance among state-of-the-art models in IMTS forecasting with low computational cost. Our code is available at https://github.com/qianlima-lab/PyOmniTS.

In this paper, a low-complexity enhanced memory polynomial search (EMPS) model is proposed for complex nonlinear memory effect scenarios in digital pre-distortion (DPD). Since the traditional polynomial models cannot describe the complex nonlinear memory effect of power amplifiers (PAs) well, this paper extends the memory polynomial basis functions to enhance modeling capabilities. Combining the basis function multiplexing strategy and the iterative greedy search method, the proposed model can achieve acceptable linearization performance with low complexity. Moreover, operations such as amplitude segmentation function or phase compensation can be introduced to extend the proposed model and increase the modeling capability. Experimental results show that the proposed model not only performs well in fully sampled wideband scenarios but also is able to obtain better DPD results in undersampled wideband scenarios.

Automatic pain assessment in postoperative children is crucial for monitoring their health and preventing potential complications. However, the automatic pain assessment still faces the following challenges. Firstly, the individual variation of painful expressions in children enhances the difficulty of mapping diverse features of expressions to pain scores accurately. Secondly, the imbalanced label distribution caused by abundant non-painful samples usually makes the model more likely to predict an unexpectedly lower pain score. To address the above challenges, we propose a novel multi-modal multi-expert framework, namelyMMF, for postoperative pain assessment in children. Specifically, the samples are clustered in each modality to train multiple expert models, each focusing on a smaller feature subspace for easier regression of pain scores. Meanwhile, some expert models are allocated to rare painful samples to relieve the side effects caused by the imbalanced distribution of labels. Moreover, a confidence-based integration of multi-modal features from multiple experts is made to achieve a more accurate final prediction. Experimental results show thatMMFexhibits superior accuracy of pain assessment on the multi-modal pain database collected from postoperative children by us. In particular,MMFcan achieve the mean absolute error (MAE) of 1.03 and the Pearson correlation coefficient (PCC) of 0.88.

As data continues to grow exponentially, the importance of online learning across various domains has increased significantly. However, most existing studies assume that the true class label for each incoming data point is readily available, an assumption that is often impractical. To address this issue, this paper introduces a novel algorithm called Density-based Clustering and One-Class Ensemble Active Learning (DCOE-AL). This approach constructs an ensemble model by combining a clustering algorithm with the One-Class Broad Learning System (OCBLS), which represents clusters within the feature space. Furthermore, a new active learning mechanism is developed to enable DCOE-AL to effectively handle challenges associated with concept drift and label scarcity. The proposed method is evaluated on multiple synthetic data streams that exhibit diverse types of concept drift, as well as on several real-world data streams. Comparative evaluations demonstrate that DCOE-AL achieves superior performance while requiring significantly fewer labeled samples.

The development of the Industrial Internet will generate a large amount of valuable data, known as industrial big data (IBD). By mining and utilizing IBD, enterprises can improve production efficiency, reduce costs and risks, optimize management processes, and innovate services and business models. However, industrial big data comes from various institutions in all walks of life and has features such as multi-source, heterogeneity, and multi-modality. And data sharing and trading (DS&T) occur in the Industrial Internet environment without mutual trust. These characteristics pose new challenges to analytics methods and privacy and security protection technologies. Therefore, this paper aims to provide references for privacy-preserving and secure industrial big data analytics (IBDA) from three perspectives: research framework, platform architecture, and key technologies. Firstly, we review the current state of research on theories and technologies related to IBDA. Then, we reveal three challenges to secure and efficient IBDA. We take the analytics and utilization of IBD as systematic engineering, propose the research framework for privacy-preserving and secure IBDA, and point out the specific content to be studied. Further, we design the architecture of the IBDA platform with the idea of layering, including a function model, security architecture, and system architecture. Finally, detailed research proposals and potential technologies for IBD analytics and utilization are presented from three aspects: data fusion and analytics, data privacy and security protection, and blockchain.

Soft skills, such as personality characteristics, communication skills and leadership, affect personal career performance greatly. Therefore, predicting the soft skills of interviewees can provide interviewers with a strong reference for the decision of hiring. Nowadays, as asynchronous video interviews have gradually become a popular form of interviews, automatic interview evaluation of soft skills has attracted widespread attention from researchers. However, existing automatic evaluation methods have two significant drawbacks. Firstly, most of them model the problem as multi-modal fusion of long-term sequences, while ignoring the consistency of multi-modal expression in short-time windows, which is a key attribute of the interview scene. Secondly, without embedding of professional knowledge in the interview field, the interpretability of the model is relatively weak. To address the above problems, we propose a novelMulti-modal Window-Consistency Fusionnetwork, namely MWCF, to capture the expression consistency of different modalities in a short-time window and re-weight the language signals to enhance important portions in verbal clues. Meanwhile, in order to enhance the interpretability of the evaluation model, we introduce the professional knowledge of interviewers by proposing a topic generation module based on question attention, and embedding the most representative keywords under different soft skills into the model. Furthermore, a real-world interview dataset is built by developing an asynchronous interview platform, and extensive experiments are conducted to show the superior performance of our proposed model.

The remote heart rate measurement aims to use the remote photoplethysmography (rPPG) technology to extract the heart rate information from face videos in a non-contact manner. Most of existing methods treat each part of the face image equally and usually have significant performance drop in some real-world application scenarios, where the face area is partially masked or occluded. To address this challenge, we propose a part-based resilient remote heart rate measurement algorithm, namely PBHeart, which is based on the fine-grained estimation of the effectiveness of different parts in a face. In particular, we apply the 3D convolutional neural network to process face frames and evenly divide the feature map into adjacent parts, each of which is used to predict the rPPG signal and the heart rate jointly. Meanwhile, we measure the effectiveness of each part of face based on the precision of prediction and the cluster based reliability of model. Finally, the prediction results from each part of face is weighted by the effectiveness, and integrated to achieve the final result. Comprehensive experiments based on three real-world datasets are conducted to prove the state-of-the-art performance of PBHeart, especially for partially masked faces. More importantly, we have collected and will release a new masked face dataset namely FMD, which includes more than 1 hour videos from 22 volunteers, to support further research in this field.

As a challenging task, unsupervised person re-identification (Re-ID) aims to optimize the pedestrian matching model based on the unlabeled image frames from surveillance videos. Recently, the fusion with the spatio-temporal clues of pedestrians have been proven effective to improve the performance of classification. However, most of these methods adopt some hard combination approaches by multiplying the visual scores with the spatio-temporal scores, which are sensitive to the noise caused by imprecise estimation of the spatio-temporal patterns in unlabeled datasets and limit the advantage of the fusion model. In this paper, we propose a Graph based Spatio-Temporal Fusion model for high-performance multi-modal person Re-ID, namely G-Fusion, to mitigate the impact of noise. In particular, we construct a graph of pedestrian images by selecting neighboring nodes based on the visual information and the transition time between cameras. Then we use a randomly initialized two-layer GraphSAGE model to obtain the multi-modal affinity matrix between images, and deploy the distillation learning to optimize the visual model by learning the affinity between the nodes. Finally, a graph-based multi-modal re-ranking method is deployed to make the decision in the testing phase for precise person Re-ID. Comprehensive experiments are conducted on two large-scale Re-ID datasets, and the results show that our method achieves a significant improvement of the performance while combined with SOTA unsupervised person Re-ID methods. Specifically, the mAP scores can reach 92.2%, and 80.4% on the Market-1501, and MSMT17 datasets respectively.

In recent years, deep learning has been widely used in remote sensing, especially in the field of synthetic aperture radar (SAR) image target detection. However, all of these deep learning models continue increasing the network’s depth and width without maintaining a good balance between accuracy and speed. Therefore, in this article, we propose a hybrid representation learning-enhanced SAR target detection algorithm based on the unique features of SAR images from a lightweight perspective called HRLE-SARDet. First, we design a lightweight and scattering feature extraction backbone that is more suitable for SAR image data. Second, for the multiscale feature discrepancy, we design a new multiscale feature fusion neck. Next, to better extract the scattering information from small targets of SAR images and improve the detection accuracy, we design a lightweight hybrid representation learning enhancement module. Finally, to better fit target detection for SAR image datasets, we redesign a more flexible loss function, which allows for an easy adjustment of the importance of polynomial bases according to the target task and dataset. Extensive experimental results on three SAR image ship target datasets (SSDD, AIR-SARShip-2.0, and HRSID) and a newly released large multiclass target SAR dataset (MSAR-1.0) show that our HRLE-SARDet achieves 98.4%, 79.2%, 92.5%, and 88.4% mean average precision (mAP) with only 1.09 M parameters and 2.5 G floating-point operations (FLOPs) on the SSDD, AIR-SARShip-2.0, HRSID, and MSAR-1.0 datasets, respectively, which is an excellent performance.

According to the surveys of the World Health Organization, distracted driving is one of main causes of road traffic accidents. To improve road traffic safety, real-time detection of drivers’ driving behavior is very important for the development of highly reliable Advanced Driver Assistance System (ADAS). At present, the deep learning architecture based on a Convolutional Neural Network (CNN) has disadvantages such as large number of parameters and weak global feature extraction ability. Therefore, this paper proposes an innovative driver distraction detection model based on the fusion of a transformer and a CNN, referred to as FPT, which is the first exploration in the field of driver distraction detection. First, we introduce the latest Twins transformer as a benchmark. Then, we design residual embedding to replace block embedding, which can further integrate the convolutional neural network with Transformer and improve the feature extraction ability. In addition, the Multilayer Perceptron (MLP) module with a large parameter occupancy rate in the original transformer structure is replaced with a lightweight group convolution module to reduce computational complexity. Finally, a cross-entropy loss function for label smoothing is designed to guide network learning with significantly differentiated features. Comparison results on two large-scale driver distraction detection datasets show that the proposed FPT offers a better compromise between computational cost and performance compared to the state-of-the-art CNN and Transformer architectures.

Deep neural networks are vulnerable to backdoor attacks where adversaries inject the trigger into partial training data to manipulate the trained model misclassification. In addition, the poisoned model behaves normally on clean inputs, and the malicious behavior only occurs when the secret trigger is present, making backdoor attacks hard to be detected. Most existing input detection methods leverage the link between triggers and outputs to reveal the poisoned inputs, which suffer from the trigger-size or the “all-to-all” attack scenario. We show that the internal activations produced by benign and poisoned inputs are significantly different in the poisoned model. In this paper, we propose a novel and run-time input detection algorithm, Activation Subset Scanning (ACTSS), which extracts the activations of incoming inputs and leverages an anomaly detection algorithm to identify malicious inputs. We search and score for the abnormal activation subset according to the statistical difference of activations between benign and poisoned data using nonparametric statistics technology. Extensive experiments are conducted on three public datasets: CIFAR10, GTSRB, and ImageNet, with three representative models. The results verify our approach's effectiveness and state-of-the-art performance, which achieve over 98% false rejection rate for different types of triggers.

Traffic accidents caused by distracted drivers account for a large proportion of traffic accidents each year, and monitoring the driving state of drivers to avoid traffic accidents caused by distracted driving has become a very important research direction. At present, the field of driver distraction detection mainly adopts supervised learning methods, which have problems such as poor generalization ability, large labeling cost, and weak artificial intelligence. This paper is oriented toward driver distraction fine-grained detection and innovatively proposes a new unsupervised deep learning algorithm, which is referred to as UDL, to achieve a more human-like level of intelligence. First, we build a new unsupervised deep learning algorithm; furthermore, we integrate the multilayer perceptron (MLP) architecture to build a new backbone and projection head to strengthen feature extraction capabilities; and finally, a new loss function based on contrast learning and a stop-gradient strategy is designed to guide the model to learn more robust features. The comparison results on large-scale driver distraction detection datasets show that our UDL method can accurately detect driver distraction without labels and exhibits excellent generalization performance with a linear evaluation accuracy of 97.38%; In addition, after fine-tuning with fewer labels, our UDL method can achieve superior performance close to state-of-the-art supervised learning methods, achieving 99.07% accuracy after fine-tuning using only 50% of the labeled data, which greatly reduces the cost and limitations of manual annotation.

Network representation learning (NRL) has far-reaching effects on data mining research, showing its importance in many real-world applications. NRL, also known as network embedding, aims at preserving graph structures in a low-dimensional space. These learned representations can be used for subsequent machine learning tasks, such as vertex classification, link prediction, and data visualization. Recently, graph convolutional network (GCN)-based models, e.g., GraphSAGE, have drawn a lot of attention for their success in inductive NRL. When conducting unsupervised learning on large-scale graphs, some of these models employ negative sampling (NS) for optimization, which encourages a target vertex to be close to its neighbors while being far from its negative samples. However, NS draws negative vertices through a random pattern or based on the degrees of vertices. Thus, the generated samples could be either highly relevant or completely unrelated to the target vertex. Moreover, as the training goes, the gradient of NS objective calculated with the inner product of the unrelated negative samples and the target vertex may become zero, which will lead to learning inferior representations. To address these problems, we propose an adversarial training method tailored for unsupervised inductive NRL on large networks. For efficiently keeping track of high-quality negative samples, we design a caching scheme with sampling and updating strategies that has a wide exploration of vertex proximity while considering training costs. Besides, the proposed method is adaptive to various existing GCN-based models without significantly complicating their optimization process. Extensive experiments show that our proposed method can achieve better performance compared with the state-of-the-art models.

Workflow scheduling problem (WSP) is a well-known combinatorial optimization problem, which is defined to assign a series of interconnected tasks to the available resources to meet user defined Quality of Service (QoS). The guided random search methods and heuristic based methods are two most common methods for solving WSP. However, these methods either require expensive computational cost or heavily rely on human's empirical knowledge, which makes them inconvenient for practical applications. Keeping this in mind, this paper proposes a cooperative coevolution hyper-heuristic framework to solve WSP with an objective of minimizing the completed time of workflow. In particular, in the proposed framework, two heuristic rules, namely, the task selection rule (TSR) and the resource selection rule (RSR), are learned automatically by a cooperative coevolution genetic programming (CCGP) algorithm. The TSR is used to select a ready task for scheduling, while the RSR is used to allocate resources to perform the selected task. To improve the search efficiency, a set of low-level heuristics are defined and used as building blocks to construct the TSR and RSR. Further, to validate the effectiveness of the proposed framework, randomly generated workflow instances and four real-world workflows are used as test cases in the experimental study. Compared with several state-of-the-art methods, e.g., the Heterogeneous Earliest Finish Time (HEFT) and the Predict Earliest Finish Time (PEFT), the high-level heuristics found by our proposed framework demonstrate superior performance on all the test cases in terms of several metrics including the schedule length ratio, speedup and efficiency.

Lip reading aims to predict the spoken sentences from silent lip videos. Due to the fact that such a vision task usually performs worse than its counterpart speech recognition, one potential scheme is to distill knowledge from a teacher pretrained by audio signals. However, the latent domain gap between the cross-modal data could lead to a learning ambiguity and thus limits the performance of lip reading. In this paper, we propose a novel collaborative framework for lip reading, and two aspects of issues are considered: 1) the teacher should understand bi-modal knowledge to possibly bridge the inherent cross-modal gap; 2) the teacher should adjust teaching contents adaptively with the evolution of the student. To these ends, we introduce a trainable "master" network which ingests both audio signals and silent lip videos instead of a pretrained teacher. The master produces logits from three modalities of features: audio modality, video modality, and their combination. To further provide an interactive strategy to fuse these knowledge organically, we regularize the master with the task-specific feedback from the student, in which the requirement of the student is implicitly embedded. Meanwhile, we involve a couple of "tutor" networks into our system as guidance for emphasizing the fruitful knowledge flexibly. In addition, we incorporate a curriculum learning design to ensure a better convergence. Extensive experiments demonstrate that the proposed network outperforms the state-of-the-art methods on several benchmarks, including in both word-level and sentence-level scenarios.

Most of the existing unsupervised cross-domain person re-identification (re-ID) methods utilize pseudo-labels estimation to cast the unsupervised problem into a supervised problem, whose performance is limited by the quality of pseudo-labels. To address the problem, we propose a noisy pseudo-label processing (NPLP) approach to suppress the pseudo-labels noise and improve the performance of the person re-ID model. Specifically, we first summarize two types of pseudo-label noise that lead to the collapse of the re-ID model, as defined as mixed noise and fragmented noise. Secondly, we propose a different method which is composed of Startup Stage and Correcting Stage for pseudo-labels estimation to relieve these two types of noise respectively. The Startup Stage aims to decrease the ratio of the fragmented noise by increasing the recall of the clustering results. At the Correcting Stage, we evaluate the quality of the pseudo-labels and correct those low-quality pseudo-labels to suppress the mixed noise and generate more reliable pseudo-labels for the re-ID model to learn. At last, we build a feature learning strategy for unsupervised re-ID task and learn from the denoised pseudo-labels iteratively. Extensive evaluations on three large-scale benchmarks show that the NPLP is competitive with most state-of-the-art unsupervised re-ID methods.

Traditional person re-identification aims to retrieve the surveillance images containing the same pedestrian. As the quick development of modern cities, a large number of multimodal personal information, including mobile information and social network log, is available and useful for customer identification and crime tracking. Compared with traditional person reidentification (ReID) only based on image modality, how to make full use of multi-modal information for efficient person ReID is more challenging. In this paper, we propose a brand new cross-modal pedestrian retrieval task based on a novel multi-modal dataset containing GPS trajectories and surveillance images. Three sub-tasks are evolved in the benchmark: unsupervised GPS-to-Image, Image-to-GPS, and Image-to-Image retrieval. In order to further verify our ideas, we propose the Similarity Driven Model (SDM), which utilizes the attention mechanism to improve the performance of domain adaptation. Furthermore, we build a cross-modal heterogeneous graph based on SDM, and adopt Triplet-Walk to uniformly represent different modalities for retrieval. Experimental results demonstrate that our method achieves the state-of-the-art on the GPS-ReID dataset.

Vehicle Re-Identification (Re-ID) is a challenging task, which aims to match the surveillance images containing the same vehicle. Since vehicles of the same type tend to share very similar appearance, slight difference in local areas are usually critical in the vehicle Re-ID task. Recently, some fine-grained Re-ID algorithms have achieved superior performance by modeling the key areas with specific semantics such as windows, lights, car orientation, etc. However, such methods are labor-intensive to label the key areas for object detection. This work proposes a Cascaded Hierarchical Context-Aware scheme namely CHCA, which is free of fine-grained labeling, to adaptively extract the visual features of discriminative local areas based on surrounding hierarchical context information with a specially designed recursive cross-level attention mechanism. It does not require any additional supervision and is easy to be embedded in existing networks. Extensive experiments on three popular vehicle Re-ID benchmarks demonstrate the effectiveness of CHCA, which has competitive results with existing state-of-the-art methods applying fine-grained labels.

Directly deploying a trained multi-modal classifier to a new environment usually leads to poor performance due to the well-known domain shift problem. Existing multi-modal domain adaptation methods treated each modality equally and optimize the sub-models of different modalities synchronously. However, as observed in this paper, the degrees of domain shift in different modalities are usually diverse. We propose a novel Differentiated Learning framework to make use of the diversity between multiple modalities for more effective domain adaptation. Specifically, we model the classifiers of different modalities as a group of teacher/student sub-models, and a novel Prototype based Reliability Measurement is presented to estimate the reliability of the recognition results made by each sub-model on the target domain. More reliable results are then picked up as teaching materials for all sub-models in the group. Considering the diversity of different modalities, each sub-model performs the Asynchronous Curriculum Learning by choosing the teaching materials from easy to hard measured by itself. Furthermore, a reliability-aware fusion scheme is proposed to combine all optimized sub-models to support final decision. Comprehensive experiments based on three multi-modal datasets with different learning tasks have been conducted, which show the superior performance of our model while comparing with state-of-the-art multi-modal domain adaptation models.

As a policy-based routing protocol, the primary purpose of Border Gateway Protocol (BGP) is to exchange routing reachability information to provide sufficient end-to-end Quality-of-Service (QoS). The constant increase of anomalous traffic of BGP affects the connectivity and reachability of routing information among different Autonomous Systems (ASs), which calls for building accurate alerting models to provide stable routing services in the Internet. The previous works classify anomalies without considering the characteristic of multiple time scales, which may lead to inaccurate classification. In this paper, we propose a novel Multi-Scale Long Short-Term Memory (MSLSTM) model to capture the anomalous behaviors from BGP traffic. In our model, a Discrete Wavelet Transform is used to obtain temporal information on multiple scales, and a hierarchical two-layer LSTM architecture is devised where the first layer learns the attentions of different time scales to generate an integrated historical representation, and the second layer captures the temporal dependency in the learned representation. To evaluate the feasibility in different alerting scenarios, we conduct comprehensive experiments based on several BGP data sets collected from real world applications. The results demonstrate that our model achieves a promising performance compared with the state-of-the-art approaches.

Person Re-Identification (Re-ID) aims to match the persons contained in surveillance videos, and is usually run on powerful servers in a supervised mode. However, centralized processing of massive video from thousands of cameras in a city is very costly and causes serious problems of privacy protection. Moreover, the labeling of numerous data for supervised training is also infeasible in this scenario. To address this problem, we propose a novel Self-optimizing Memory Network model, namely SoMem, which runs person Re-ID on edge devices in a totally unsupervised and distributed way. Specifically, SoMem adopts a random walk based collaborative training procedure to optimize the visual model on each camera based on locally collected images, and builds a distributed memory network to memorize and match the observed persons by using a distributed mutual ranking algorithm. Based on the cross-camera person matching results learned by the memory network, the visual models on edge devices are further optimized in a self-organized manner. Comprehensive experiments are conducted on several real person Re-ID datasets and deployed on edge devices to show the effectiveness and efficiency of this novel distributed Re-ID model.

Multi-domain event detection (MED) task aims to discover real-world events from multi-domain and multi-view data sources. Most of the existing methods process multiple data modalities alone or concatenate single modality for the MED task, ignoring diverse intrinsic structures and underlying relations between each modality. In this paper, we address the heterogeneity of modalities' structure and data views' relations, and present a novel heterogeneous graph embedding method HGE2MED to learn the heterogeneous underlying relations and robust representation of each node. In particular, we construct heterogeneous graph for multi-domain and multi-view data, and obtain heterogeneous walking sequences by heterogeneous random walk. Using our heterogeneous graph embedding method to sample the triplets from above sequences, we are able to obtain robust embedding of each node by optimizing the network. Finally, taking the embedding of news as input, the event labels are predicted by a simple linear classifier. Comprehensive experiments are conducted to demonstrate the effectiveness of our proposed approach over state-of-the-art event detection algorithms on a real-world dataset.

Internet platforms provide new ways for people to share experiences, generating massive amounts of data related to various real-world concepts. In this paper, we present an event detection framework to discover real-world events from multiple data domains, including online news media and social media. As multi-domain data possess multiple data views that are heterogeneous, initial dictionaries consisting of labeled data samples are exploited to align the multi-view data. Furthermore, a shared multi-view data representation (SMDR) model is devised, which learns underlying and intrinsic structures shared among the data views by considering the structures underlying the data, data variations, and informativeness of dictionaries. SMDR incorpvarious constraints in the objective function, including shared representation, low-rank, local invariance, reconstruction error, and dictionary independence constraints. Given the data representations achieved by SMDR, class-wise residual models are designed to discover the events underlying the data based on the reconstruction residuals. Extensive experiments conducted on two real-world event detection datasets, i.e., Multi-domain and Multi-modality Event Detection dataset, and MediaEval Social Event Detection 2014 dataset, indicating the effectiveness of the proposed approaches.

Precise destination prediction from partial trajectories have a huge potential impact on intelligent location-based approaches. Traditional prediction approaches, which treat trajectories as one-dimensional sequences and process them in a single scale, fail to capture diverse and rich two-dimensional patterns of trajectories in different spatial scales. Meanwhile, most models treat each portion of a trajectory equally in terms of contributing to final destination prediction. This is in conflict with our observation that there exists some albeit small local areas playing much more important roles for destination prediction than the others. To address these problems, we propose a novel prediction algorithmT-CONV, which models trajectories as two-dimensional images, and then feed them into a convolutional neural network (CNN) architecture to extract multi-scale patterns for precise destination prediction. Furthermore, we propose a method to extract regions with different relevance for final output ofT-CONV, and further explore the local patterns of important regions by integrating multi-scope local-enhancement areas based on attention mechanism. The comprehensive experiments based on two large-scale real taxi trajectory datasets show thatT-CONVcan achieve higher accuracy than the state-of-the-art methods, demonstrating the strength of the multi-scale and multi-scope feature extraction mechanisms in trajectory mining.

In this article, we propose to learn shared semantic space with correlation alignment ( S 3 CA ) for multimodal data representations, which aligns nonlinear correlations of multimodal data distributions in deep neural networks designed for heterogeneous data. In the context of cross-modal (event) retrieval, we design a neural network with convolutional layers and fully connected layers to extract features for images, including images on Flickr-like social media. Simultaneously, we exploit a fully connected neural network to extract semantic features for text documents, including news articles from news media. In particular, nonlinear correlations of layer activations in the two neural networks are aligned with correlation alignment during the joint training of the networks. Furthermore, we project the multimodal data into a shared semantic space for cross-modal (event) retrieval, where the distances between heterogeneous data samples can be measured directly. In addition, we contribute a Wiki-Flickr Event dataset, where the multimodal data samples are not describing each other in pairs like the existing paired datasets, but all of them are describing semantic events. Extensive experiments conducted on both paired and unpaired datasets manifest the effectiveness of S 3 CA , outperforming the state-of-the-art methods.

Scholars' homepages are important places to show personal research interest and academic achievement through the Web. However, according to our observation, only a small portion of scholars update their publications and related events on their homepages in time. In this paper, we propose a homepage augmentation technique, which automatically shows the newest academic events related to a scholar on his/her homepage. Specifically, we model the relations between homepages and the events collected from the Web as a complex heterogenous network, and propose an Embedding-based Heterogenous random Walk algorithm, namely EHWalk, to predict the links between homepages and events. Compared with existing embedding-based link prediction algorithms, EHWalk supports more efficient modeling of complex heterogenous relations in a dynamically changing network, which helps link the massive new updated events to homepages precisely and efficiently. Comprehensive experiments on a real-world dataset are conducted and the results show that our algorithm can achieve both good effectiveness and efficiency for real-world deployment.

Most of the proposed person re-identification algorithms conduct supervised training and testing on single labeled datasets with small size, so directly deploying these trained models to a large-scale real-world camera network may lead to poor performance due to underfitting. It is challenging to incrementally optimize the models by using the abundant unlabeled data collected from the target domain. To address this challenge, we propose an unsupervised incremental learning algorithm, TFusion, which is aided by the transfer learning of the pedestrians' spatio-temporal patterns in the target domain. Specifically, the algorithm firstly transfers the visual classifier trained from small labeled source dataset to the unlabeled target dataset so as to learn the pedestrians' spatial-temporal patterns. Secondly, a Bayesian fusion model is proposed to combine the learned spatio-temporal patterns with visual features to achieve a significantly improved classifier. Finally, we propose a learning-to-rank based mutual promotion procedure to incrementally optimize the classifiers based on the unlabeled data in the target domain. Comprehensive experiments based on multiple real surveillance datasets are conducted, and the results show that our algorithm gains significant improvement compared with the state-of-art cross-dataset unsupervised person re-identification algorithms.

Detecting anomalous Border Gateway Protocol (BGP) traffic is significantly important in improving both security and robustness of the Internet. Existing solutions apply classic classifiers to make real-time decision based on the traffic features of present moment. However, due to the frequently happening burst and noise in dynamic Internet traffic, the decision based on short-term features is not reliable. To address this problem, we propose MS-LSTM, a multi-scale Long Short-Term Memory (LSTM) model to consider the Internet flow as a multi-dimensional time sequence and learn the traffic pattern from historical features in a sliding time window. In addition, we find that adopting different time scale to preprocess the traffic flow has great impact on the performance of all classifiers. In this paper, comprehensive experiments are conducted and the results show that a proper time scale can improve about 10% accuracy of LSTM as well as all conventional machine learning methods. Particularly, MS-LSTM with optimal time scale 8 can achieve 99.5% accuracy in the best case.

Supporting random jump in P2P-VoD systems requires efficient lookup for the “best” suppliers, where “best” means the suppliers should meet two requirements: content match and network quality match . Most studies use a DHT-based method to provide content lookup; however, these methods are neither able to meet the network quality requirements nor suitable for VoD streaming due to the large overhead. In this paper, we propose Mediacoop, a novel hierarchical lookup scheme combining both content and quality match to provide random jumps for P2P-VoD systems. It exploits the play position to efficiently locate the candidate suppliers with required data (content match), and performs refined lookup within the candidates to meet quality match. Theoretical analysis and simulation results show that Mediacoop is able to achieve lower jump latency and control overhead than the typical DHT-based method. Moreover, we implement Mediacoop in a BitTorrent-like P2P-VoD system called CoolFish and make optimizations for such “total cache” applications. The implementation and evaluation in CoolFish show that Mediacoop is able to improve user experiences, especially the jump latency, which verifies the practicability of our design.

The random seeking in P2P-VoD system requires efficient lookup for ¿good¿ suppliers. The main challenge is that good suppliers should meet two requirements: ¿content match¿ and ¿quality match¿, while most existing methods only focus on one aspect. In this paper, we propose Mediacoop, a novel structured lookup method combining both content and quality match to provide random seeking for P2P-VoD services. It exploits playpoint distance to efficiently locate the candidate suppliers with required data (content match), and performs refined lookup within the candidates to meet quality match. Theoretical analysis and simulations show that Mediacoop outperforms the traditional methods. Our real-world system also proves the effectiveness of the design.

Inspired by how search behavior works in human society, we propose CTO, a self-organized semantic overlay based on concept tree for P2P IR infrastructure, which is efficient for full text search in pure P2P environment without any central control or powerful peer as hub node. Especially, CTO performs very well on searching the unpopular resources shared by a few peers. In our experiment, while searching for the scarce documents shared by the peers, CTO achieves about 80% recall rate when the search covers less than 5% peers in the overlay. The search latency of CTO is also very low, which is controlled in the range about 5~12 hops.

Large-scale distributed systems, with thousands or even millions of nodes, produce complex and dynamic behaviors. Packet-level simulation is necessary to test and analyze these systems, such as grids, peer-to-peer (P2P) applications as well as worm and DDoS containment systems. However, the current network simulators are not convenient for application layer simulation. We present the NSME, a visualized parallel simulator that allows researchers to simulate their applications in a large virtual network with most details transparent to them. A hierarchical routing is used to enhance the fidelity of simulation and reduce the memory cost of network construction. A variation of CMB algorithm is implemented for parallel synchronization. We show the functions of NSME by three applications: 1) a self-similar background traffic model, 2) a Slammer worm spreading model, and 3) a P2P live streaming system, which demonstrate its effectiveness for simulating any large-scale distributed applications. The performance tests show that the memory cost of NSME is distinctly lower than NS2 and the parallel efficiency can reach about 60% in any of the above applications.

BitTorrent (BT) is one of the most popular Peer-to- Peer (P2P) protocols for delivering media files in the Internet today. Although BT is quite efficient for sharing and downloading files by using P2P swarming technique, the users have to download almost the whole media file before playing it. This is determined by the Rarest-Block-Download-First strategy of standard BT implementations, which is designed for fast delivery of files in the systems but not for streaming application. In this paper, we present LiveBT, a new protocol which supports video-on- demand streaming service and is totally compatible to the current BitTorrent protocol. LiveBT enables users to play hot movies shared in the BT systems smoothly just after 2~3 minutes of buffering time. We also develop the prototype of LiveBT and test the performance through the real BT download tasks of media files. By comparing our prototype with some popular BT clients claiming to support view-as- download service such as Bitcomet, we find that LiveBT spends a much shorter buffering time to play and achieves quite smooth playback performance.

Compared with centralized information retrieval system, peer-to-peer (P2P) full text information retrieval system is more scalable, cost effective, fault tolerant, and powerful to search scarce content at the edge of Internet. However, P2P full text search is a very challenging problem, and the traditional broadcast ways are quite ineffective. We present a P2P full text information retrieval system WonGoo that is based on P2P structured overlay networks. We also implement a simulation program and test the performance of the system. Experimental results show that WonGoo is a steady system with high recall, good load balance and low resource usage.