Federated learning (FL) is a promising technology for data privacy and distributed optimization, but it suffers from data imbalance and heterogeneity among clients. Existing FL methods try to solve the problems by aligning client with server model or by correcting client model with control variables. These methods excel on IID and general Non-IID data but perform mediocrely in Simpson's Paradox scenarios. Simpson's Paradox refers to the phenomenon that the trend observed on the global dataset disappears or reverses on a subset, which may lead to the fact that global model obtained through aggregation in FL does not accurately reflect the distribution of global data. Thus, we propose FedCFA, an novel FL framework employing counterfactual learning to generate counterfactual samples by replacing local data critical factors with global average data, aligning local data distributions with the global and mitigating Simpson's Paradox effects. In addition, to improve the counterfactual samples quality, we introduce factor decorrelation (FDC) loss to reduce the correlation among features and thus improve the independence of extracted factors. We conduct extensive experiments on six datasets and verify that our method outperforms other FL methods in terms of efficiency and global model accuracy under limited communication rounds.

With the proliferation of mobile devices and IoT applications, edge caching has become vital for mitigating network congestion and enhancing user Quality of Experience (QoE). However, traditional caching policies, such as Least Frequently Used (LFU), First-In-First-Out (FIFO), and Least Recently Used (LRU), often struggle to perform effectively in highly dynamic and heterogeneous environments, particularly when content sizes vary significantly. Moreover, existing approaches, whether AI-driven or heuristic-based, typically adopt a single caching strategy, which inherently limits their flexibility and adaptability. To address these limitations, we propose ECMSA, a learning-based multi-strategy edge caching algorithm that integrates a reinforcement learning-driven proactive caching strategy with three conventional reactive caching strategies. Specifically, ECMSA operates in two stages: First, it generates four candidate cache lists—three derived from traditional caching policies (LFU, FIFO, LRU) and one produced by our self-attention-enhanced Deep Deterministic Policy Gradient (Atten-Actor DDPG)-based proactive caching strategy. Next, it employs another Atten-Actor DDPG agent to dynamically select the optimal strategy in real time, leveraging current state features. This dual-agent framework enables continuous learning and adaptation of caching decisions, effectively optimizing content placement and update policies in response to evolving user demands. Extensive experiments conducted on both synthetic and real-world datasets demonstrate that ECMSA achieves 15-17% higher cache-hit ratios and 16-22% lower latency than baseline methods under constrained cache capacities and diverse content sizes. Furthermore, ECMSA exhibits strong robustness and generalization ability, allowing it to rapidly adapt to unseen environments.

Federated Learning (FL) enables multiple clients to collaboratively train models without sharing raw data, making it well-suited for privacy-preserving applications in heterogeneous IoT environments. However, disparities in client model architectures and computational resources often lead to accuracy degradation and the straggler problem, undermining training efficiency. To address these challenges, we propose HAPFL, a novel Heterogeneity-aware Personalized Federated Learning framework based on multi-level Reinforcement Learning (RL). HAPFL integrates three key components: 1) An RL-based model allocation mechanism that employs a PPO agent to assign appropriately sized models to clients based on their computing capabilities; 2) An RL-based training intensity adjustment scheme that dynamically controls local training epochs per client to reduce straggling latency; 3) A mutual learning scheme using knowledge distillation between each client's local model and a homogeneous lightweight model (LiteModel), which also serves as the global aggregation model to tackle model heterogeneity. Experiments on MNIST, CIFAR-10, and ImageNet-10 demonstrate that HAPFL achieves superior accuracy while reducing overall training time by$20.9 \%-40.4 \%$and straggling latency by$\mathbf{1 9. 0 \% - 4 8. 0 \%}$compared to existing approaches.

Multiaccess edge computing (MEC) has emerged as a promising solution that can enable low-end terminal devices to run large complex applications by offloading their tasks to edge servers. The task offloading strategy, determining how to offload tasks, remains the most critical issue of MEC. Traditional offloading approaches either suffer from high computational complexity or poor self-adjustability to dynamic changes in the edge environment. Deep reinforcement learning (DRL) provides an effective way to tackle these issues. However, most existing DRL-based methods solely consider either a continuous or a discrete action space, where the limited action space results in accuracy loss and restricts the optimality of offloading decisions. Nevertheless, the edge task offloading problem in practice often confronts both discrete and continuous actions. In this article, we propose a tailored proximal policy optimization (PPO)-based method, named Hybrid-PPO, enhanced by the parameterized discrete-continuous hybrid action space. Assisted with Hybrid-PPO, we further design a novel DRL-based multiserver multitask collaborative partial task offloading scheme adhering to a series of specifically built formal models. Experimental results prove that our approach achieves high offloading efficiency and outperforms the existing state-of-the-art offloading schemes in terms of convergence rate, energy cost, time cost, and generalizability under various network conditions.

Abstract Due to the emergence of graph convolutional networks (GCNs), the skeleton-based action recognition has achieved remarkable results. However, the current models for skeleton-based action analysis treat skeleton sequences as a series of graphs, aggregating features of the entire sequence by alternately extracting spatial and temporal features, i.e., using a 2D (spatial features) plus 1D (temporal features) approach for feature extraction. This undoubtedly overlooks the complex spatiotemporal fusion relationships between joints during motion, making it challenging for models to capture the connections between different temporal frames and joints. In this paper, we propose a Multimodal Graph Self-Attention Network (MGSAN), which combines GCNs with self-attention to model the spatiotemporal relationships between skeleton sequences. Firstly, we design graph self-attention (GSA) blocks to capture the intrinsic topology and long-term temporal dependencies between joints. Secondly, we propose a multi-scale spatio-temporal convolutional network for channel-wise topology modeling (CW-TCN) to model short-term smooth temporal information of joint movements. Finally, we propose a multimodal fusion strategy to fuse joint, joint movement, and bone flow, providing the model with a richer set of multimodal features to make better predictions. The proposed MGSAN achieves state-of-the-art performance on three large-scale skeleton-based action recognition datasets, with accuracy of 93.1% on NTU RGB+D 60 cross-subject benchmark, 90.3% on NTU RGB+D 120 cross-subject benchmark, and 97.0% on the NW-UCLA dataset. Code is available at https://github.com/lizaowo/MGSAN .

With the ever-increasing demand for computing power in deep learning, distributed training techniques have proven to be effective in meeting these demands. However, current existing state-of-the-art distributed training frameworks, such as Parameter Server (PS), Ring-All-Reduce, and their varieties, still face significant challenges. In particular, the existence of communication bottlenecks can severely limit the efficiency and scalability of distributed training frameworks, making it difficult to fully and effectively exert the computing power of large-scale clusters, especially in the presence of dynamic and ever-changing network environments. To address these issues and further maximize the utilization of the computing power of clusters, in this paper we propose an efficient and dynamic distributed training framework, named GreedW. GreedW can greatly improve the training efficiency of workers by dynamically constructing an adaptive customized communication network and adaptively scheduling the workload. Specifically, GreedW employs a greedy strategy to dynamically construct the communication network tree in each iteration for gradient transmission with minimum communication cost and applies a heterogeneity-aware workload allocation scheme to adaptively balance the heavy traffic across heterogeneous workers in the cluster taking into account the available computing capabilities of each node, which effectively alleviates the network bottleneck. It is worth noting that GreedW is enabled to dynamically adjust the assigned job on each worker node based on their completion time during each round of model aggregation to ensure that each worker node completes its assignments around the same time, thus mitigating the intractable straggler issue and minimizing their idle waiting time. Comprehensive experimental evaluations on three different-scaled training models (i.e., Mnist-2NN, Mnist-CNN, and TextCNN) for image recognition and natural language processing tasks demonstrate that GreedW outperforms the existing state-of-the-art frameworks in terms of training efficiency, system flexibility, and robustness.

Efficient one-hop multicasting (OHM) of high-volume sensor data plays a pivotal role in the success of cooperative autonomous driving applications. Although millimeter-Wave (mmWave) bands demonstrate huge potential for high- bandwidth OHM data transmission, the challenge lies in enabling individual vehicles to locate and communicate with suitable neighbors in a fully distributed and highly dynamic scenario. This paper introduces mmV2V, a fully distributed OHM scheme designed for vehicular networks, comprising three tightly integrated protocols. Initially, synchronized vehicles perform a probabilistic neighbor discovery procedure, wherein randomly divided transmitters (or receivers) clockwise scan (or listen to) the surroundings in synchronization with heterogeneous Tx (or Rx) beams. This approach facilitates the identification of the vast majority of neighbors within a few repeated rounds. Subsequently, vehicles engage in negotiations with their neighbors to establish an optimal communication schedule in evenly distributed slots. Finally, matched pairs of neighboring vehicles commence high data rate transmissions using refined beams. We implement a prototype testbed to validate the feasibility of the main components of mmV2V. Extensive simulations based on generated and real-world traffic traces are conducted and the results demonstrate that mmV2V consistently achieves a high completion ratio in demanding OHM tasks across various traffic conditions.

To tackle the issue of heterogeneous input raw data samples obtained by different devices and enhance the feature extraction capability of edge devices, we propose a vertically split neural network based edge-device collaborative artificial intelligence (AI) inference framework. The local results calculated by various light-size sub-networks at edge devices are transmitted and aggregated at the server for the downstream inference task. Nevertheless, the transmission of such high-dimensional local results involves severe communication overhead. To resolve this issue, the technique of over-the-air computation (AirComp) is adopted to enable low-latency aggregation. The same entry of all devices’ local results is transmitted over a same wireless resource block and aggregated via the waveform superposition property. Furthermore, to simultaneously support the aggregation of all dimensions of the local results, we consider a broadband channel and leverage orthogonal frequency division multiplexing (OFDM) to divide the system bandwidth into multiple subcarriers which are then assigned for different dimensions. Consequently, an extra degree of freedom is introduced to design the aggregation of all dimensions. We then propose a scheme of joint subcarrier allocation, power allocation, and receiver beamforming to minimize the aggregation distortion and enhance inference performance. Extensive experiments are conducted to verify the superiority of the proposed design over benchmarks.

Visual question answering (VQA) is a fundamental and challenging cross-modal task. This task requires the model to fully understand the image’s content and reason out the answer based on the question. Existing VQA models understand visual content mainly based on bottom-up or grid features. However, both types of vision features have some drawbacks. The discreteness and independence of bottom-up features pre-vent models from adequately performing relational reasoning. Image segmentation by grid features leads to the fragmentation of meaningful visual regions, limiting the cross-modal alignment capability of the model. Therefore, we proposed a more flexible method called Visual Graph. It can connect different patches according to semantic similarity and spatial relevance to model the potential relationships and cluster the adjacent homologous patches. Based on the Visual Graph, we designed a Visual Graph Reasoning Network for VQA. We evaluated our model on GQA and VQA-v2. The experimental results show that our models can achieve excellent performance between single models.

As a critical concern of multi-access edge computing (MEC), task offloading has received extensive attention. Although deep reinforcement learning (DRL) has achieved great success in resolving the task offloading problem, most existing DRL-based offloading schemes only consider either continuous action space or discrete action space, which results in the loss of optimality of decisions. Moreover, the generalization ability of the existing schemes is still far from adaptive to dynamic changes in the environment. This leads to offloading strategies having to conduct re-sampling and re-training, which largely impairs the offloading efficiency. To address these issues, we propose a novel efficient MEC task offloading scheme based on parameterized meta-reinforcement learning taking hybrid action space into account. We first formulate this problem as a non-convex multi-objective optimization problem. Then, we design a parameterized meta-reinforcement learning algorithm, named Meta-Hybrid-PPO, with hybrid action space to solve the optimization problem. Comprehensive experimental results show that our Meta-Hybrid-PPO not only performs better than existing state-of-the-art methods in reducing task processing latency and computational energy consumption but also achieves better adaptability.

To ensure the delivery of high-performance and reliable services, data center networks (DCNs) are often over-provisioned for peak workload and traffic bursts. However, in real-world data centers, network traffic seldom reaches peak capacity of the network, resulting in significant energy waste. Traditional energy conservation approaches either suffer from high computational complexity and low solution quality, or their strategies cannot be dynamically adjusted to accommodate changes in data center network traffic. Deep reinforcement learning (DRL) provides an effective way to deal with these issues. However, most of the existing DRL-based schemes only consider either a continuous action space or a discrete action space, which greatly restricts the optimality of decisions. To solve these problems, this paper proposes a novel DRL-based DCN energy optimization framework, named SmartDCN. Specifically, SmartDCN consists of a traffic prediction module (TPM) and an energy optimization module (EOM). TPM incorporates an improved LSTM model JANET with an attention mechanism providing a high prediction accuracy, while EOM integrates our newly proposed parameterized DRL algorithm , named PAS-DQN, combining with the discrete-continuous hybrid action space. PAS-DQN implements a two-level control mechanism for the network, using TPM to predict future traffic in the data center as input. It is devoted to dynamically aggregating current traffic and makes tradeoffs between energy efficiency, performance, and robustness to optimize the network’s power consumption by dynamically calculating the minimum required network subset and turning off the non-involved network devices to achieve power savings. Experimental results show that SmartDCN significantly outperforms the existing state-of-the-art schemes in terms of energy savings under various network conditions.

Interactive medical image segmentation based on human-in-the-loop machine learning is a novel paradigm that draws on human expert knowledge to assist medical image segmentation. However, existing methods often fall into what we call interactive misunderstanding, the essence of which is the dilemma in trading off short- and long-term interaction information. To better use the interaction information at various timescales, we propose an interactive segmentation framework, called interactive MEdical image segmentation with self-adaptive Confidence CAlibration (MECCA), which combines action-based confidence learning and multi-agent reinforcement learning. A novel confidence network is learned by predicting the alignment level of the action with short-term interaction information. A confidence-based reward-shaping mechanism is then proposed to explicitly incorporate confidence in the policy gradient calculation, thus directly correcting the model’s interactive misunderstanding. MECCA also enables user-friendly interactions by reducing the interaction intensity and difficulty via label generation and interaction guidance, respectively. Numerical experiments on different segmentation tasks show that MECCA can significantly improve short- and long-term interaction information utilization efficiency with remarkably fewer labeled samples. The demo video is available at https://bit.ly/mecca-demo-video .

As an alternative to the deep learning model, deep forest outperforms deep neural networks in many aspects with fewer hyperparameters and better robustness. To improve the computing performance of deep forest, ForestLayer proposes an efficient task-parallel algorithm S-FTA at a fine sub-forest granularity, but the granularity of the sub-forest cannot be adaptively adjusted. BLB-gcForest further proposes an adaptive sub-forest splitting algorithm to dynamically adjust the sub-forest granularity. However, with distributed storage, its BLB method needs to scan the whole dataset when sampling, which generates considerable communication overhead. Moreover, BLB-gcForest's tree-based vector aggregation produces extensive redundant transfers and significantly degrades the system's performance in vector aggregation stage. To deal with these existing issues and further improve the computing efficiency and scalability of the distributed deep forest, in this paper, we propose a novel Computing-Efficient and RobusT distributed Deep Forest framework, named CERT-DF. CERT-DF integrates three customized schemes, namely, block-level pre-sampling, two-stage pre-aggregation, and system-level backup. Specifically, CERT-DF adopts the block-level pre-sampling method to implement data blocks' local sampling eliminating frequent data remote access and maximizing parallel efficiency, applies the two-stage pre-aggregation method to adjust the class vector aggregation granularity to greatly decrease the communication overhead, and leverages the system-level backup method to enhance the system's disaster tolerance and immensely accelerate task recovery with minimal system resource overhead. Comprehensive experimental evaluations on multiple datasets show that our CERT-DF significantly outperforms the state-of-the-art approaches with higher computing efficiency, lower system resource overhead, and better system robustness while ensuring good accuracy.

Deformable registration of multi-modal medical images is a challenging task in medical image processing due to the differences in both appearance and structure. We propose an unsupervised hierarchical translation-based model to perform a coarse to fine registration of multi-modal medical images. The proposed model consists of three parts: a coarse registration network, a modal translation network and a fine registration network. First, the coarse registration network learns to obtain the coarse deformation field, which is applied as structure-preserving information to generate a translated image by the modal translation network. Then, the translated image as enhancing information combined with the original images are used to derive a fine deformation field in the fine registration network. Furthermore, the final deformation field is composed from the coarse and the fine deformation fields. In this way, the proposed model can learn high accurate deformation field to implement multi-modal medical image registration. Experiments on two multi-modal brain image datasets demonstrate the effectiveness of this model.

Recently, several Transformer-based methods have been presented to improve image segmentation. However, since Transformer needs regular square images and has difficulty in obtaining local feature information, the performance of image segmentation is seriously affected. In this paper, we propose a novel encoder-decoder network named TCRNet, which makes Transformer, Convolutional neural network (CNN) and Recurrent neural network (RNN) complement each other. In the encoder, we extract and concatenate the feature maps from Transformer and CNN to effectively capture global and local feature information of images. Then in the decoder, we utilize convolutional RNN in the proposed recurrent decoding unit to refine the feature maps from the decoder for finer prediction. Experimental results on three medical datasets demonstrate that TCRNet effectively improves the segmentation precision.

One-hop multicasting (OHM) of high-volume sensor data is essential for cooperative autonomous driving applications. While millimeter-Wave (mmWave) bands can be utilized for high-bandwidth OHM data transmission, it is very challenging for individual vehicles to find and communicate with a proper neighbor in a fully distributed and highly dynamic scenario. In this paper, we propose a fully distributed OHM scheme in vehicular networks, called mmV2V, which consists of three highly integrated protocols. Specifically, synchronized vehicles first conduct a probabilistic neighbor discovery procedure, in which randomly divided transmitters (or receivers) clockwise scan (or listen to) the surroundings in pace with heterogeneous Tx (or Rx) beams. In this way, the vast majority of neighbors can be identified in a few repeated rounds. Furthermore, vehicles negotiate with each of their neighbors about the optimal communication schedule in evenly distributed slots. Finally, each agreed pair of neighboring vehicles start high data rate transmissions with refined beams. We conduct extensive simulations and the results demonstrate that mmV2V can achieve a high completion ratio in rigid OHM tasks under various traffic conditions.

As an efficient technique for resource sharing in data centers, network virtualization enables resource multiplexing by allowing multiple heterogeneous virtual networks (VNs) to simultaneously coexist on the shared substrate infrastructure. How to effectively embed the VNs onto the substrate network is known as the virtual network embedding (VNE) problem. However, as an NP-hard problem, the VNE problem-solving suffers a high computation complexity. Artificial Intelligence (AI) provides a promising way to alleviate these issues. However, the existing AI-based works still cannot fully and efficiently leverage substrate network information to formulate embedding policies. To this end, in this paper we propose a novel deep reinforcement learning (DRL) based coordinated VNE algorithm, called Intelligent Coordinated Embedding (ICE). To reduce the computation complexity, ICE adopts an efficient resource abstraction model, Blocking Island (BI), which greatly reduces the search space. With the benefit of DRL and BI, ICE can efficiently adjust embedding strategies according to the environment states, aiming to maximize resource utilization and overall revenue while minimizing the embedding cost. The experimental results prove that ICE outperforms both the traditional non-DRL-based approach and the state-of-the-art DRL-based approach.

Graph convolutional networks (GCNs) are widely used in skeleton-based action recognition and have achieved excellent results. However, it is evident that the convolution operation can lead to losing some original input information. The incomplete utilisation of original input data limits GCNs’ ability to obtain the skeleton’s correlation. This paper proposes a graph instinctive attention convolutional network (GIAN) to solve this problem. In particular, it contains an instinctive attention module that uses self-attention to obtain the correlation within the original input skeleton. Then, parameter attention is used to further refine the relationship between different skeleton joints. Experimental results on publicly available datasets demonstrate that the GIAN outperforms most of the state-of-the-art algorithms.

When a detection model that has been well-trained on a set of classes faces new classes, incremental learning is always necessary to adapt the model to detect the new classes. In most scenarios, it is required to preserve the learned knowledge of the old classes during incremental learning rather than reusing the training data from the old classes. Since the objects in remote sensing images often appear in various sizes, arbitrary directions, and dense distribution, it further makes incremental learning-based object detection more difficult. In this article, a new architecture for incremental object detection is proposed based on feature pyramid and knowledge distillation. Especially, by means of a feature pyramid network (FPN), the objects with various scales are detected in the different layers of the feature pyramid. Motivated by Learning without Forgetting (LwF), a new branch is expended in the last layer of FPN, and knowledge distillation is applied to the outputs of the old branch to maintain the old learning capability for the old classes. Multitask learning is adopted to jointly optimize the losses from two branches. Experiments on two widely used remote sensing data sets show our promising performance compared with state-of-the-art incremental object detection methods.

As an emulous alternative to deep neural networks, Deep Forest emerges with features like low complexity, fewer hyper-parameters, and good robustness, which are predominantly desired in distributed computing applications and ecosystems. Recently, an efficient distributed Deep Forest system, named ForestLayer, was proposed, designing a fine-grained sub-Forest-based task-parallel algorithm to improve the parallel computing efficiency of Deep Forest. However, the sub-Forest splitting of ForestLayer is static and one-off without adaptability to the computing environment, nevertheless, the size of splitting granularity has a significant impact on the system performance. To further improve the computing efficiency and scalability of the distributed Deep Forest, in this paper, we propose a novel distributed Deep Forest algorithm, named BLB-gcForest (Bag of Little Bootstraps-gcForest), which augments the gcForest (multi-Grained Cascade Forest) approach for constructing Deep Forest. BLB-gcForest carries out parallel computation for each tree in sub-Forests at a finer parallel granularity and integrates with the Bag of Little Bootstraps (BLB) mechanism to reduce massive transmitted feature instances for Cascade Forest Layers, utterly improving both computation efficiency and communication efficiency. Moreover, to solve the problem of the forest splitting granularity, we further design an adaptive sub-Forest splitting algorithm to ensure the maximum resource utilization for parallel computation of each sub-Forest. Experimental results on four well-known large-scale datasets, namely YEAST, LETTER, MNIST, CIFAR10, show that the training efficiency of BLB-gcForest achieves up to 20.3x and 1.64x speedups compared with the state-of-the-art gcForest and ForestLayer, respectively while guaranteeing higher accuracy and better robustness

Prediction tasks about students such as predicting students' academic performances have practical real-world significance at both the student level and the college level. With the rapid construction of smart campuses, colleges not only offer residence and academic programs but also record students' daily life. The digital footprints provide an opportunity to offer better solutions for prediction tasks. In this paper, we aim to propose a general deep neural network which can jointly model student heterogeneous daily behaviors generated from digital footprints and social influence to deal with prediction tasks. To this end, we design a variant of LSTM and a novel attention mechanism to model the daily behavior sequence. The proposed LSTM is able to consider context information (e.g., weather conditions) while modeling the daily behavior sequence. The proposed attention mechanism can dynamically learn the different importance degrees of different days for every student. Based on behavior information, we propose an unsupervised way to construct a social network to model social influence. Moreover, we design a residual network based decoder to model the complex interactions between the features and get the predicted values such as future academic performances. Qualitative and quantitative experiments on two real-world datasets collected from a college have demonstrated the effectiveness of our model.

As a supervised learning method, Factorization Machine (FM) is famous for its capability of modeling feature interactions. However, FM's performance might be bad if we assign the same weight to all feature interactions, as not all of them are equally useful and productive. Attentional Factorization Machine (AFM) improves FM by discriminating the importance of distinctive feature interactions via a neural attention network. Nevertheless, the neural attention network in AFM is not fine-grained enough and it ignores the information of the fields implied by the features, which limits the performance of the model. In this work, we propose a novel model named IO-aware Factorization Machine (IOFM), which enhances the feature representation ability of attention mechanism in estimating weights via two awareness auxiliary matrices. To make the model more efficient, we further reduce the model parameters using canonical decomposition for the two auxiliary matrices and design a shared matrix to correlate the decomposed matrices. Extensive experiments on two real-world datasets indicate the superiority of our IOFM model over the state-of-the-art methods.

Images taken under different camera poses are rotated or distorted, which leads to poor perception experiences. This article proposes a new framework to automatically transform the images to the conformable view setting by assembling different convolution neural networks. Specifically, a referential three-dimensional ground plane is first derived from the color image and a novel projection mapping algorithm is developed to achieve automatic viewing transformation. Extensive experimental results demonstrate that the proposed method outperforms the state-of-the-art vanishing points based methods by a large margin in terms of accuracy and robustness.

Nonrigid image registration plays an important role in the field of computer vision and medical application. The methods based on Demons algorithm for image registration usually use intensity difference as similarity criteria. However, intensity based methods can not preserve image texture details well and are limited by local minima. In order to solve these problems, we propose a Gabor feature based LogDemons registration method in this paper, called GFDemons. We extract Gabor features of the registered images to construct feature similarity metric since Gabor filters are suitable to extract image texture information. Furthermore, because of the weak gradients in some image regions, the update fields are too small to transform the moving image to the fixed image correctly. In order to compensate this deficiency, we propose an inertial constraint strategy based on GFDemons, named IGFDemons, using the previous update fields to provide guided information for the current update field. The inertial constraint strategy can further improve the performance of the proposed method in terms of accuracy and convergence. We conduct experiments on three different types of images and the results demonstrate that the proposed methods achieve better performance than some popular methods.

Predicting the accurate traffic crowd flows is of practical importance for intelligent transportation systems (ITS). However, it is challenging because traffic flows are affected by multiple complex factors, such as spatial and temporal dependencies of regions and external factors. In this paper, we propose a deep hybrid spatio-temporal dynamic neural network, called DHSTNet, to predict both inflows and outflows in every region of a city. More specifically, it consists of four main components, i.e., closeness influence taking the instantaneous variations of traffic flows, period influence regularly identifying daily changes of traffic crowd flows, weekly component identifying the patterns of weekly traffic flows and external component acquiring external factors. We design a branch of deep hybrid recurrent convolutional neural network units to model the first three temporal properties, i.e., closeness, period influence, and weekly influence. The external components are feed into two fully connected neural networks. For different branches, our proposed model assigns different weights and then combines the output of the four components. Experimental results based on two large-scale real-world datasets demonstrate the superiority of our model over the existing state-of-the-art methods.

Human-human interaction recognition has attracted increasing attention in recent years due to its wide applications in computer vision fields. Currently there are few publicly available RGBD-based human-human interaction datasets collected. This paper introduces a new dataset for human-human interaction recognition. Furthermore, a novel feature descriptor based on spatial relationship and semantic motion trend similarity between body parts is proposed for human-human interaction recognition. The motion trend of each skeleton joint is firstly quantified into the specific semantic word and then a Kernel is built for measuring the similarity of either intra or inter body parts by histogram interaction. Finally, the proposed feature descriptor is evaluated on the SBU interaction dataset and the collected dataset. Experimental results demonstrate the outperformance of our method over the state-of-the-art methods.

In this paper, we present an effective and accurate gaze estimation method based on two-eye model of a subject with the tolerance of free head movement from a Kinect sensor. To accurately and efficiently determine the point of gaze, i) we employ two-eye model to improve the estimation accuracy; ii) we propose an improved convolution-based means of gradients method to localize the iris center in 3D space; iii) we present a new personal calibration method that only needs one calibration point. The method approximates the visual axis as a line from the iris center to the gaze point to determine the eyeball centers and the Kappa angles. The final point of gaze can be calculated by using the calibrated personal eye parameters. We experimentally evaluate the proposed gaze estimation method on eleven subjects. Experimental results demonstrate that our gaze estimation method has an average estimation accuracy around 1.99°, which outperforms many leading methods in the state-of-the-art.

Motor vehicle theft has caused massive economic loss over the world. This paper proposes an embedded vision system to detect automotive interior intrusion. The system uses a fusion of an acceleration module and a vision module to meet the requirement of low power consumption for most motor vehicles. Furthermore, an effective intrusion detection algorithm is developed for the on-board vision module. The vision system is able to detect the intrusion even in the dark night due to the employment of infrared lights. Experimental evaluation is conducted under a variety of illumination conditions, such as day time, night time and even shining light. An intrusion detection accuracy of 91.7% is achieved, which shows that the developed embedded vision system is reliable for motor vehicle intrusion detection.

In recent years, with the popularity of smart mobile devices, mobile Internet has rapidly developed. When the social network meets the localization technology, it gives birth to a Location-Based Social Network (LBSN). The situational awareness based on the location has become more research significance. However, how to combine the context awareness, mobile sensors and abundant users' historical location data to make the platform more efficient, how to ensure that the proceeds can improve the accuracy of recommendation and perceived task assignment, are still challenges in the location-based social network. In this paper, we demonstrate a model to use historical location data of the participants and analyze the point-of-interest (POI). Then we propose the user location and the empirical value algorithm PTHS based on the HITS algorithm. Through analyzing the interest points of the selected scene perception task, we find that those users have similar point-of-interest, and rank them by the location and experience PTHS algorithm. Finally, much more appropriate users are assigned to the tasks. Through theoretical analysis and extensive simulations, we validate that proposed method is effective and efficient.

With widespread usages of smart phones, participatory sensing becomes mainstream, especially for applications requiring pervasive deployments with massive sensors. However, the sensors on smart phones are prone to the unknown measurement errors, requiring automatic calibration among uncooperative participants. Current methods need either collaboration or explicit calibration process. However, due to the uncooperative and uncontrollable nature of the participants, these methods fail to calibrate sensor nodes effectively. We investigate sensor calibration in monitoring pollution sources, without explicit calibration process in uncooperative environment. We leverage the opportunity in sensing diversity, where a participant will sense multiple pollution sources when roaming in the area. Further, inspired by expectation maximization (EM) method, we propose a two-level iterative algorithm to estimate the source presences, source parameters and sensor noise iteratively. The key insight is that, only based on the participatory observations, we can “calibrate sensors without explicit or cooperative calibrating process”. Theoretical analysis proves that, our method can converge to the optimal estimation of sensor noise, where the likelihood of observations is maximized. Also, extensive simulations show that, ours improves the estimation accuracy of sensor bias up to 20 percent and that of sensor noise deviation up to 30 percent, compared with three baseline methods.

Wireless LANs, particularly WiFi, have been pervasively deployed and have fostered myriad wireless communication services and ubiquitous computing applications. A primary concern in designing these applications is to combat harsh indoor propagation environments, particularly Non-Line-Of-Sight (NLOS) propagation. The ability to identify the existence of the Line-Of-Sight (LOS) path acts as a key enabler for adaptive communication, cognitive radios, and robust localization. Enabling such capability on commodity WiFi infrastructure, however, is prohibitive due to the coarse multipath resolution with MAC-layer received signal strength. In this paper, we propose two PHY-layer channel-statistics-based features from both the time and frequency domains. To further break away from the intrinsic bandwidth limit of WiFi, we extend to the spatial domain and harness natural mobility to magnify the randomness of NLOS paths while retaining the deterministic nature of the LOS component. We propose LiFi, a statistical LOS identification scheme with commodity WiFi infrastructure, and evaluate it in typical indoor environments covering an area of 1500 m2. Experimental results demonstrate that LiFi achieves an overall LOS detection rate of 90.42% with a false alarm rate of 9.34% for the temporal feature and an overall LOS detection rate of 93.09% with a false alarm rate of 7.29% for the spectral feature.

This paper proposes a novel method with a modified Expectation-Maximisation (EM) Algorithm to segment hand gestures in the RGB-D data captured by Kinect. With the depth map and RGB image aligned by the genetic algorithm to estimate the key points from both depth and RGB images, a novel approach is proposed to refine the edge of the tracked hand gesture, which is used to segment the RGB image of the hand gestures, by applying a modified EM algorithm based on Bayesian networks. The experimental results demonstrated the modified EM algorithm effectively adjusts the RGB edges of the segmented hand gestures. The proposed methods have potential to improve the performance of hand gesture recognition in Human-Computer Interaction (HCI).

VCPS (Vehicular Cyber Physical Systems) is a special kind of networked cyber physical system in which each vehicle is regarded as a communication unit. Vehicle's movement is restricted by road and environment in VCPS, while traditional random mobility model and waypoint mobility model cannot reflect the realistic vehicle traces. In VCPS, with the high speed of vehicles, the network topology undergoing tremendous changes all the time, which greatly undermines the stability of communication between vehicles. The diversity and complexity of traffic scenarios in VCPS have also increased the difficulty of designing an efficient and stable routing protocol. In this paper, we creatively combine SDN (Software Defined Networking) and VCPS together and propose a new VCPS communication architecture, which enable VCPS to be manageable by remote controller. SD-VCPS can flexibly change routing policies depending on different traffic scenes or traffic periods, adjusting the topology of VCPS to adapt to different network requirements. We further present a new location-based routing protocol for SD-VCPS, and corroborate the efficiency of our proposed framework by experiments using network simulator NS3.

The information field is undergoing a new round of technological revolution from the Internet to the Internet of things. Vehicle ad-hoc network (VANET), an application of the internet of things using in Intelligent Transportation System (ITS), has already attracted broad attention in the world in recent years. It mainly provides communications between vehicle-to-vehicle and vehicle-to-infrastructure, which significantly improve road transport efficiency, reduce energy consumption and ease traffic congestion. In this paper, we developed a client to make SUMO and NS3 work parallel by TraCI (Traffic Control Interface) in NS3. It helps NS3 get SUMO's information and sends instructions to change the states of vehicles and traffic lights. We present a realistic road traffic model with kinds of vehicles and intelligent traffic lights. The model is built in SUMO (Simulation of Urban Mobility). We use OpenStreetMap to generate a realistic map near the bund in Shanghai. The traffic flow is built according to a survey which makes us get meaningful and reliable statistics. A mechanism of changing the traffic lights dynamically is introduced to minimize traffic jams and give high priority to emergency vehicle. As a result, the waiting time and the duration of the vehicles in the scenario have reduced significantly after using the mechanism. The emergency vehicle's waiting time is less than others.

Off-the-shelf smartphones have boosted large scale participatory sensing applications as they are equipped with various functional sensors, possess powerful computation and communication capabilities, and proliferate at a breathtaking pace. Yet the low participation level of smartphone users due to various resource consumptions, such as time and power, remains a hurdle that prevents the enjoyment brought by sensing applications. Recently, some researchers have done pioneer works in motivating users to contribute their resources by designing incentive mechanisms, which are able to provide certain rewards for participation. However, none of these works considered smartphone users' nature of opportunistically occurring in the area of interest. Specifically, for a general smartphone sensing application, the platform would distribute tasks to each user on her arrival and has to make an immediate decision according to the user's reply. To accommodate this general setting, we design three online incentive mechanisms, named TBA, TOIM and TOIMAD, based on online reverse auction. TBA is designed to pursue platform utility maximization, while TOIM and TOIM-AD achieve the crucial property of truthfulness. All mechanisms possess the desired properties of computational efficiency, individual rationality, and profitability. Besides, they are highly competitive compared to the optimal offline solution. The extensive simulation results reveal the impact of the key parameters and show good approximation to the state-of-the-art offline mechanism.

The standard back-propagation(BP) algorithm converges slowly and is easy to trap into local minimum, which are the main reasons why it cannot be used widely in real-time applications. Therefore, a novel BP algorithm based on three-term method consisting of a learning rate (LR), a momentum factor (MF) and a proportional factor (PF), called the TTMBP algorithm, was put forward in this paper. The convergence speed and stability were enhanced by adding PF. The self-adapting learning and self-adjusting-architecture methods are adopted in order that a moderate size networks model can be obtained according to environmental requirements. The novel BP algorithm is proposed to solve the problem of service selection in ubiquitous computing. We have fulfilled simulation in an actual power supply system for communication devices and the results of simulation show that the proposed control scheme is not only scalable but also efficient. The control scheme based on novel BP algorithm superior to the traditional service selection method based on trust mechanism. It can exactly choose a most suitable service from many target services and give the most perfect service performance to users