While gradient aggregation playing a vital role in federated or collaborative learning, recent studies have revealed that gradient aggregation may suffer from some attacks, such as gradient inversion, where the private training data can be recovered from the shared gradients. However, the performance of the existing attack methods is limited because they usually require prior knowledge in Batch Normalization and could only reconstruct a single image or a small batch one. To make the attacks less restrictive and more applicable, we propose an effective and practical gradient inversion method in this paper. Specifically, we use cosine similarity to measure the difference of gradients between the synthesized and ground-truth images, and then construct an input regularization for the fully connected layer to ensure the fidelity of the image. Moreover, we apply the total variation denoising strategy to the convolution feature map for further improving the smoothness of the reconstructed image. Experimental results demonstrate that our method can reconstruct high fidelity training data on a large batch size for complex data sets, such as ImageNet.
Abstract. With the purpose of predicting icing of transmission lines, a model updating approach is presented in this study. The changes of structural dynamic response of transmission lines that is caused by icing is studied firstly by using finite element method. Then, model updating method and particle swarm optimization is implemented to indentify the thickness of icing according to the alternation of natural frequencies. The results show that the proposed methodology is meaningful to monitor line icing. IntroductionIcing of transmission lines would cause line galloping, insulator flashover and even tower collapse which is a great threat for the operation of power grid [1-2]. Thus, effective monitoring and condition assessment of line icing is the key issue which needs to be solved, and the establishment of icing forecast plays an important role in the operation of power grid.It's very difficult to evaluate the condition of lines manually considering that transmission lines are located sparsely. Hence, the on-line monitoring technique has attracted increasing attentions in engineering application. The current monitoring strategy are mainly based on measuring the weight, angle of insulator, wind speed, temperature and humidity. The equivalent thickness of line icing can be estimated according to measured data [3][4][5], based on that, related staff will be alerted as long as the thickness evaluated beyond a predefined value. However, the actual icing condition is usually distributed non-uniformly along the length of line, thus, the monitoring technique based on the evaluation of equivalent thickness is unable to get more information of icing distribution in details. In this work, a methodology for icing forecasting based on model updating method is presented. The modal frequencies are introduced to identify the icing condition of transmission lines with the help of artificial intelligence technique, and the results show that the proposed approach is able to identify and predict the distribution of icing.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.