News recommendation aiming to find attractive news for users has been received many attentions in recent years. Existing news recommendation methods mainly focus on modeling user preference based on the interaction be...
详细信息
Submarines are vital for military defense and national security in strategic communication systems. Power Spectral Density (PSD) characterizes signal spectrum frequency dependence and is crucial for analyzing broadban...
详细信息
With the development of the architectures and the growth of AIoT application requirements, data processing on edge has become popular. Neural network inference is widely employed for data analytics on edge devices. Th...
With the development of the architectures and the growth of AIoT application requirements, data processing on edge has become popular. Neural network inference is widely employed for data analytics on edge devices. This paper extensively explores neural network inference on integrated edge devices and proposes EdgeNN, the first neural network inference solution on CPU-GPU integrated edge devices. EdgeNN has three novel characteristics. First, EdgeNN can adaptively utilize the unified physical memory and conduct the zero-copy optimization. Second, EdgeNN involves a novel inference-targeted inter- and intra-kernel CPU-GPU hybrid execution approach, which co-runs the CPU with the GPU to fully utilize the edge device’s computing resources. Third, EdgeNN adopts a fine-grained adaptive inference tuning approach, which can divide the complicated inference structure into sub-tasks mapped to the CPU and the GPU. Experiments show that on six popular neural network inference tasks, EdgeNN brings an average of 3.97×, 3.12×, and 8.80× speedups to inference on the CPU of the integrated device, inference on a mobile phone CPU, and inference on an edge CPU device. Additionally, it achieves 22.02% time benefits to the direct execution of the original programs. Specifically, 9.93% comes from better utilization of unified memory, and 10.76% comes from CPU-GPU hybrid execution. Besides, EdgeNN can deliver 29.14× and 5.70× higher energy efficiency than the edge CPU and the discrete GPU, respectively. We have made EdgeNN available at https://***/ChenyangZhang-cs/EdgeNN.
In recent years, the Internet of Things (IoT) technology has developedby leaps and bounds. However, the large and heterogeneous networkstructure of IoT brings high management costs. In particular, the low costof IoT d...
详细信息
In recent years, the Internet of Things (IoT) technology has developedby leaps and bounds. However, the large and heterogeneous networkstructure of IoT brings high management costs. In particular, the low costof IoT devices exposes them to more serious security concerns. First, aconvolutional neural network intrusion detection system for IoT devices isproposed. After cleaning and preprocessing the NSL-KDD dataset, this paperuses feature engineering methods to select appropriate features. Then, basedon the combination of DCNN and machine learning, this paper designs acloud-based loss function, which adopts a regularization method to preventoverfitting. The model consists of one input layer, two convolutional layers,two pooling layers and three fully connected layers and one output ***, a framework that can fully consider the user’s privacy protection isproposed. The framework can only exchange model parameters or intermediateresults without exchanging local individuals or sample data. This paperfurther builds a global model based on virtual fusion data, so as to achievea balance between data privacy protection and data sharing computing. Theperformance indicators such as accuracy, precision, recall, F1 score, and AUCof the model are verified by simulation. The results show that the model ishelpful in solving the problem that the IoT intrusion detection system cannotachieve high precision and low cost at the same time.
Hierarchical multi-granularity classification is the task of classifying objects according to multiple levels or granularities. The class hierarchy is vital side information for hierarchical multi-granularity classifi...
详细信息
Traffic prediction is indispensable for constructing transportation networks in smart cities. Due to the complex spatio-temporal correlations of traffic data, this task presents challenges. Recent studies mainly use g...
详细信息
Image classification is challenging due to the high dimensionality and large variations of the image data. As an effective method in image classification, feature learning can be considered a multiobjective problem of...
详细信息
The causality relation modeling remains a challenging task for group activity recognition. The causality relations describe the influence on the centric actor (effect actor) from its correlative actors (cause actors)....
The causality relation modeling remains a challenging task for group activity recognition. The causality relations describe the influence on the centric actor (effect actor) from its correlative actors (cause actors). Most existing graph models focus on learning the actor relation with synchronous temporal features, which is insufficient to deal with the causality relation with asynchronous temporal features. In this paper, we propose an Actor-Centric Causality Graph Model, which learns the asynchronous temporal causality relation with three modules, i.e., an asynchronous temporal causality relation detection module, a causality feature fusion module, and a causality relation graph inference module. First, given a centric actor and its correlative actor, we analyze their influences to detect causality relation. We estimate the self influence of the centric actor with self regression. We estimate the correlative influence from the correlative actor to the centric actor with correlative regression, which uses asynchronous features at different timestamps. Second, we synchronize the two action features by estimating the temporal delay between the cause action and the effect action. The synchronized features are used to enhance the feature of the effect action with a channel-wise fusion. Third, we describe the nodes (actors) with causality features and learn the edges by fusing the causality relation with the appearance relation and distance relation. The causality relation graph inference provides crucial features of effect action, which are complementary to the base model using synchronous relation inference. Experiments show that our method achieves state-of-the-art performance on the Volleyball dataset and Collective Activity dataset.
Container technology is widely used and improve the efficiency of container real-time migration has become an important research topic. Existing studies mainly focus on optimizing iterative dumping of containers witho...
详细信息
With the frequent occurrence of telecommunications and networkfraud crimes in recent years, new frauds have emerged one after another whichhas caused huge losses to the people. However, due to the lack of an effective...
详细信息
With the frequent occurrence of telecommunications and networkfraud crimes in recent years, new frauds have emerged one after another whichhas caused huge losses to the people. However, due to the lack of an effectivepreventive mechanism, the police are often in a passive position. Usingtechnologies such as web crawlers, feature engineering, deep learning, andartificial intelligence, this paper proposes a user portrait fraudwarning schemebased on Weibo public data. First, we perform preliminary screening andcleaning based on the keyword “defrauded” to obtain valid fraudulent userIdentity Documents (IDs). The basic information and account information ofthese users is user-labeled to achieve the purpose of distinguishing the typesof fraud. Secondly, through feature engineering technologies such as avatarrecognition, Artificial Intelligence (AI) sentiment analysis, data screening,and follower blogger type analysis, these pictures and texts will be abstractedinto user preferences and personality characteristics which integrate multidimensionalinformation to build user portraits. Third, deep neural networktraining is performed on the cube. 80% percent of the data is predicted basedon the N-way K-shot problem and used to train the model, and the remaining20% is used for model accuracy evaluation. Experiments have shown thatFew-short learning has higher accuracy compared with Long Short TermMemory (LSTM), Recurrent Neural Networks (RNN) and ConvolutionalNeural Network (CNN). On this basis, this paper develops a WeChat smallprogram for early warning of telecommunications network fraud based onuser portraits. When the user enters some personal information on the frontend, the back-end database can perform correlation analysis by itself, so as tomatch the most likely fraud types and give relevant early warning *** fraud warning model is highly scaleable. The data of other Applications(APPs) can be extended to further improve the efficiency of anti-fraud whichhas e
暂无评论