Computing-intensive and latency-sensitive user requests pose significant challenges to traditional cloud *** response to these challenges,mobile edge computing(MEC)has emerged as a new paradigm that extends the comput...
详细信息
Computing-intensive and latency-sensitive user requests pose significant challenges to traditional cloud *** response to these challenges,mobile edge computing(MEC)has emerged as a new paradigm that extends the computational,caching,and communication capabilities of cloud *** caching certain services on edge nodes,computational support can be provided for requests that are offloaded to the ***,previous studies on task offloading have generally not considered the impact of caching mechanisms and the cache space occupied by *** oversight can lead to problems,such as high delays in task executions and invalidation of offloading *** optimize task response time and ensure the availability of task offloading decisions,we investigate a task offloading method that considers caching ***,we incorporate the cache information of MEC into the model of task offloading and reduce the task offloading problem as a mixed integer nonlinear programming(MINLP)***,we propose an integer particle swarm optimization and improved genetic algorithm(IPSO_IGA)to solve the ***_IGA exploits the evolutionary framework of particle swarm *** it uses a crossover operator to update the positions of particles and an improved mutation operator to maintain the diversity of ***,extensive simulation experiments are conducted to evaluate the performance of the proposed *** experimental results demonstrate that IPSO_IGA can save 20%to 82%of the task completion time,compared with state-of-theart and classical ***,IPSO_IGA is suitable for scenarios with complex network structures and computing-intensive tasks.
Neural code models (NCMs) have demonstrated extraordinary capabilities in code intelligence tasks. Meanwhile, the security of NCMs and NCMs-based systems has garnered increasing attention. In particular, NCMs are ofte...
详细信息
Neural code models (NCMs) have demonstrated extraordinary capabilities in code intelligence tasks. Meanwhile, the security of NCMs and NCMs-based systems has garnered increasing attention. In particular, NCMs are often trained on large-scale data from potentially untrustworthy sources, providing attackers with the opportunity to manipulate them by inserting crafted samples into the data. This type of attack is called a code poisoning attack (also known as a backdoor attack). It allows attackers to implant backdoors in NCMs and thus control model behavior, which poses a significant security threat. However, there is still a lack of effective techniques for detecting various complex code poisoning attacks. In this paper, we propose an innovative and lightweight technique for code poisoning detection named KILLBADCODE. KILLBADCODE is designed based on our insight that code poisoning disrupts the naturalness of code. Specifically, KILLBADCODE first builds a code language model (CodeLM) on a lightweight n-gram language model. Then, given poisoned data, KILLBADCODE utilizes CodeLM to identify those tokens in (poisoned) code snippets that will make the code snippets more natural after being deleted as trigger tokens. Considering that the removal of some normal tokens in a single sample might also enhance code naturalness, leading to a high false positive rate (FPR), we aggregate the cumulative improvement of each token across all samples. Finally, KILLBADCODE purifies the poisoned data by removing all poisoned samples containing the identified trigger tokens. We conduct extensive experiments to evaluate the effectiveness and efficiency of KILLBADCODE, involving two types of advanced code poisoning attacks (a total of five poisoning strategies) and datasets from four representative code intelligence tasks. The experimental results demonstrate that across 20 code poisoning detection scenarios, KILLBADCODE achieves an average FPR of 8.30% and an average Recall of 100%, signif
Accurate traffic forecasting is a critical function of intelligent transportation systems, which remains challenging due to the complex spatial and temporal dependence of traffic data. GNN-based traffic forecasting mo...
详细信息
Accurate traffic forecasting is a critical function of intelligent transportation systems, which remains challenging due to the complex spatial and temporal dependence of traffic data. GNN-based traffic forecasting models typically utilize predefined graphical structures based on prior knowledge and do not adapt well to dynamically changing traffic characteristics, which may limit their performance. The transformer is a compelling architecture with an innate global self-attention mechanism, but cannot capture low-level detail very well. In this paper, we propose a novel Spatial-Temporal Gated Hybrid Transformer Network (STGHTN), which leverages local features from temporal gated convolution, spatial gated graph convolution respectively and global features by transformer to further improve the traffic flow forecasting results. First, in the temporal dimension, we take full advantage of the local properties of temporal gated convolution and the global properties of transformer to effectively fuse short-term and long-term temporal dependence. Second, we mutually integrate two modules to complement each representation by utilizing spatial gated graph convolution to extract local spatial dependence and transformer to extract global spatial dependence. Furthermore, we propose a multi-graph model that constructs a road connection graph, a similarity graph, and an adaptive dynamic graph to exploit the static and dynamic associations between road networks. Experiments on four real datasets confirm the proposed method's state-of-the-art performance. Our implementation of the STGHTN code via PyTorch is available at hups://***/STGHTN.
In an increasingly connected digital world, the strategic use of data is critical to improving predictive accuracy, optimizing processes, and driving innovative business models. Data spaces can facilitate sovereign an...
详细信息
ISBN:
(纸本)9798400709845
In an increasingly connected digital world, the strategic use of data is critical to improving predictive accuracy, optimizing processes, and driving innovative business models. Data spaces can facilitate sovereign and secure data sharing across multiple organizations. A challenge in this environment is the lack of trust among data space participants, calling for independent data trustees. However, finding and selecting an appropriate data trustee can be difficult and is usually a manual process. Our research addresses this challenge by introducing a whitelisting approach for automatically selecting appropriate data trustees within a data space. We report on the results of an ongoing design science research project and describe the development of extensions to the Eclipse Dataspace Components, which enable the application of the whitelisting approach. These extensions are designed to support secure, efficient, and trusted data sharing. Our findings can support practitioners in the data space community in implementing robust mechanisms for selecting data trustees, thereby improving the functionality and trustworthiness of data spaces.
Convolutional neural network (CNN) has been broadly adopted on hyperspectral image (HSI) processing due to its impressive feature extraction capabilities. Nevertheless, it is still a challenge for CNN-based hyperspect...
详细信息
Large deep neural network (DNN) models have demonstrated exceptional performance across diverse downstream tasks. Sharded data parallelism (SDP) has been widely used to reduce the memory footprint of model states. In ...
详细信息
ISBN:
(数字)9798331509712
ISBN:
(纸本)9798331509729
Large deep neural network (DNN) models have demonstrated exceptional performance across diverse downstream tasks. Sharded data parallelism (SDP) has been widely used to reduce the memory footprint of model states. In a DNN training cluster, a device usually has multiple inter-device links that connect to other devices, like NVLink and InfiniBand. However, existing SDP approaches employ a single link at any given time, encountering challenges in efficient training due to significant communication overheads. We observe that the inter-device links can work independently without affecting each other. To reduce the fatal communication overhead of distributed training of large DNNs, this paper introduces HSDP, an efficient SDP training approach that enables the simultaneous utilization of multiple inter-device links. HSDP partitions models in a novel fine-grained manner and orchestrates the communication processes of partitioned parameters while considering inter-device links. This design enables concurrent communication execution and reduces communication overhead. To further optimize the training performance of HSDP, we propose a HSDP planner. The HSDP planner first abstracts the model partition and execution of HSDP into a communication parallel strategy, and builds a cost model to estimate the performance of each strategy. We then formulate the strategy searching as an optimization problem and solve it with an off-the-shelf solver. Evaluations on representative DNN workloads demonstrate that HSDP achieves up to 1.30× speedup compared to the state-of-the-art SDP training approaches.
During long-term operation,the performance of obstacles would be changed due to the material accumulating upslope the ***,the effects of retained material on impact,overflow and landing dynamics of granular flow have ...
详细信息
During long-term operation,the performance of obstacles would be changed due to the material accumulating upslope the ***,the effects of retained material on impact,overflow and landing dynamics of granular flow have not yet been *** address this gap,physical flume tests and discrete element simulations are conducted considering a range of normalized deposition height h0/H from 0 to 1,where h0 and H represent the deposition height and obstacle height,*** analytical model is modified to evaluate the flow velocity and flow depth after interacting with the retained materials,which further serve to calculate the peak impact force on the ***,the computed impact forces successfully predict the experimental results when a≥25°.In addition,the results indicate that a higher h0/H leads to a lower dynamic impact force,a greater landing distance L,and a larger landing coefficient Cr,where Cr is the ratio of slope-parallel component of landing velocity to flow velocity just before *** to the existing overflow model,the measured landing distance L is underestimated by up to 30%,and therefore it is insufficient for obstacle design when there is retained ***,the recommended Cr in current design practice is found to be nonconservative for estimating the landing velocity of geophysical *** study provides insightful scientific basis for designing obstacles with deposition.
Symbolic music generation relies on the contextual representation capabilities of the generative model, where the most prevalent approach is the Transformer-based model. Learning contextual representations are also re...
详细信息
Symbolic music generation relies on the contextual representation capabilities of the generative model, where the most prevalent approach is the Transformer-based model. Learning contextual representations are also related to the structural elements in music, i.e., intro, verse, and chorus, which have not received much attention of scientific publications. In this paper, we propose a hierarchical Transformer model to learn multiscale contexts in music. In the encoding phase, we first design a fragment scope localization module to separate the music parts into chords and sections. Then, we use a multiscale attention mechanism to learn note-, chord-, and section-level contexts. In the decoding phase, we propose a hierarchical Transformer model that uses fine decoders to generate sections in parallel and a coarse decoder to decode the combined music. We also designed a music style normalization layer to achieve a consistent music style between the generated sections. Our model is evaluated on two open MIDI datasets. Experiments show that our model outperforms other comparative models in 50% (6 out of 12 metrics) and 83.3% (10 out of 12 metrics) of the quantitative metrics for short- and long-term music generation, respectively. Preliminary visual analysis also suggests its potential in following compositional rules, such as reuse of rhythmic patterns and critical melodies, which are associated with improved music quality.
Team adaptation to new cooperative tasks is a hallmark of human intelligence, which has yet to be fully realized in learning agents. Previous studies on multi-agent transfer learning have accommodated teams of differe...
详细信息
Team adaptation to new cooperative tasks is a hallmark of human intelligence, which has yet to be fully realized in learning agents. Previous studies on multi-agent transfer learning have accommodated teams of different sizes but heavily relied on the generalization ability of neural networks for adapting to unseen tasks. We posit that the relationship among tasks provides key information for policy *** utilize this relationship for efficient transfer by attempting to discover and exploit the knowledge among tasks from different teams, proposing to learn an effect-based task representation as a common latent space among tasks, and using it to build an alternatively fixed training scheme. Herein, we demonstrate that task representation can capture the relationship among teams and generalize to unseen tasks. Thus, the proposed method helps transfer the learned cooperation knowledge to new tasks after training on a few source ***, the learned transferred policies help solve tasks that are difficult to learn from scratch.
Although the gate-diffusion input (GDI) technique supports low power and small area compared to conventional CMOS standard cell, it is limited to design larger integrated circuits for several reasons. It is difficult ...
详细信息
暂无评论