Adversarial attacks are now becoming quite a dangerous means of disrupting imageprocessing systems that use machine learning methods for decision making. Therefore, developing effective countermeasures against advers...
详细信息
The exponential growth of technological advancements in satellite and airborne remote sensing is giving rise to large volumes of high-dimensional hyperspectral image data. Apache Spark is one of the most popular, exte...
详细信息
The medical routine of the future is strongly influenced by medical information technology. The quality and efficiency of medicine are at higher standards due to the image-based methods and the increase in computation...
详细信息
Graph Neural Networks (GNNs) have emerged as potent models for graph learning. Distributing the training process across multiple computing nodes is the most promising solution to address the challenges of ever-growing...
详细信息
ISBN:
(纸本)9783031697654;9783031697661
Graph Neural Networks (GNNs) have emerged as potent models for graph learning. Distributing the training process across multiple computing nodes is the most promising solution to address the challenges of ever-growing real-world graphs. However, current adversarial attack methods on GNNs neglect the characteristics and applications of the distributed scenario, leading to suboptimal performance and inefficiency in attacking distributed GNN training. In this study, we introduce Disttack, the first framework of adversarial attacks for distributed GNN training that leverages the characteristics of frequent gradient updates in a distributed system. Specifically, Disttack corrupts distributed GNN training by injecting adversarial attacks into one single computing node. The attacked subgraphs are precisely perturbed to induce an abnormal gradient ascent in backpropagation, disrupting gradient synchronization between computing nodes and thus leading to a significant performance decline of the trained GNN. We evaluate Disttack on four large real-world graphs by attacking five widely adopted GNNs. Compared with the state-of-the-art attack method, experimental results demonstrate that Disttack amplifies the model accuracy degradation by 2.75x and achieves speedup by 17.33x on average while maintaining unnoticeability.
Federated learning (FL) is a distributed machine learning approach that reduces data transfer by aggregating gradients from multiple users. However, this process raises concerns about user privacy, leading to the emer...
详细信息
ISBN:
(纸本)9783031697654;9783031697661
Federated learning (FL) is a distributed machine learning approach that reduces data transfer by aggregating gradients from multiple users. However, this process raises concerns about user privacy, leading to the emergence of privacy preserving FL. Unfortunately, this development poses new Byzantine-robustness challenges as poisoning attacks become difficult to detect. Existing byzantine-robust algorithms operate primarily in plaintext, and crucially, current byzantine-robust privacy FL methods fail to concurrently defend against adaptive attacks. In response, we propose a lightweight, byzantine-robust, and privacy-preserving federated learning framework (LRFL), employing shuffle functions and encryption masks to ensure privacy. In addition, we comprehensively calculate the similarity of the direction and magnitude of each gradient vector to ensure byzantine-robustness. To the best of our knowledge, LRFL is the first byzantine-robust privacy preserving FL capable of identifying malicious users based on gradient angles and magnitudes. What's more, the theoretical complexity of LRFL is O(dN + dN logN), comparable to byzantine-robust FL with user number N and gradient dimension d. Experimental results demonstrate that LRFL achieves similar accuracy to state-of-the-art methods under multiple attack scenarios.
Given the computational complexity of deep neural networks (DNN), accurate prediction of their training and inference time using performance modeling is crucial for efficient infrastructure planning and DNN developmen...
详细信息
ISBN:
(纸本)9798400717932
Given the computational complexity of deep neural networks (DNN), accurate prediction of their training and inference time using performance modeling is crucial for efficient infrastructure planning and DNN development. However, existing methods often predict only the inference time and rely on exhaustive benchmarking and fine tuning, making them time consuming and restricted in scope. As a remedy, we propose ConvMeter, a novel yet simple performance model that considers the inherent characteristics of DNNs, such as architecture, dataset, and target hardware, which strongly affect their runtime and scalability. Our performance model, which has been thoroughly tested on convolutional neural networks (ConvNets), a class of DNNs widely used for image analysis, offers the prediction of inference and training time, the latter on one or more compute nodes. Experiments with various ConvNets demonstrate that our runtime predictions of inference and training phases achieved an average error rate of less than 20% and 18%, respectively, making the assessment of ConvNets regarding efficiency and scalability straightforward.
Dynamic channel pruning is a technique aimed at reducing the theoretical computational complexity and inference latency of convolutional neural networks. Dynamic channel pruning methods introduce complex additional mo...
详细信息
To address the issue of lengthy optimization processes for ATR engines using swarm intelligence optimization algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), a novel Ensemble Neural Net...
详细信息
ISBN:
(纸本)9780791887967
To address the issue of lengthy optimization processes for ATR engines using swarm intelligence optimization algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), a novel Ensemble Neural Networks based parallel Performance Optimization (ENNPPO) methodology is proposed. This strategy amalgamates several small neural networks into a comprehensive larger model for the construction of an advanced onboard ATR engine model, moreover, the independence of each small neural networks allows for the reduction of computational time through parallelprocessing on GPUs. The method then harnesses the PSO algorithm to refine performance. Numerical simulations carried out with embedded GPU computing chips demonstrate that the average optimization time per test case is 0.28 seconds, with an energy consumption of 1.22 Joules. The proposed method achieves an optimization precision that is 78.40% and 84.30% higher than that of traditional neural network approaches and Linear Parameter-Varying (LPV) model methods, respectively. Furthermore, the computation speed is enhanced by a factor of at least 7.8 times than CPU, directly attributable to the expedited parallelprocessing enabled by GPUs.
The complex glyph structures and diverse writing styles of ancient Chinese character images lead to suboptimal performance when existing image retrieval methods are directly applied to datasets of these images. Addres...
详细信息
This research aims to determine an optimal neural network model for image segmentation, addressing a crucial aspect of computer vision and deep learning. The study aims to identify high-performance neural network arch...
详细信息
ISBN:
(纸本)9798350364613;9798350364606
This research aims to determine an optimal neural network model for image segmentation, addressing a crucial aspect of computer vision and deep learning. The study aims to identify high-performance neural network architectures characterized by superior accuracy and minimized complexity through a combination of deep learning principles, multi-objective optimization and genetic algorithms. The proposed approach adapts NSGA-iii to generate new neural network architectures encoded via binary representation. The chromosomes are then decoded to undergo the training. The focus of this work is to explore and identify better models based on performance metrics: intersection over union, accuracy, and frequency weighted intersection over union (fwloU), while simultaneously minimizing model complexity by optimizing the number of parameters. The primary results are very encouraging, and in future work, we aim to provide more tests and analysis on other benchmarks.
暂无评论