检索结果-内蒙古大学图书馆

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Chu-Xiao Zuo Zhi-Jun Jia Wu-Jun Li Department of Computer Science and Technology National Key Laboratory for Novel Software Technology Nanjing University China

Speaker identification (SI) systems have been widely employed in real-world applications. However, recent research has demonstrated that SI systems are vulnerable to two prevalent attacks even without providing feedback to the attacker: the transfer-based adversarial attack and the speech synthesis spoofing attack. The transfer-based adversarial attack faces the challenges of collecting natural speech for specific content and timbre. In contrast, the speech synthesis spoofing attack can synthesize speech for any content and timbre but can be detected by audio deepfake detectors (ADD). In this paper, we propose a novel method, called adversarial text-to-speech synthesis (AdvTTS), for attacking SI systems. AdvTTS combines the strengths of transfer-based adversarial attacks and speech synthesis spoofing attacks, by synthesizing transferable attack speech with local surrogate models. AdvTTS is the first attack method that can conduct both adversarial and spoofing attacks with any speech content and timbre. AdvTTS can deceive SI systems with high-quality speech while evading ADD detection. Experiments show that AdvTTS can outperform other baselines for spoofing attacks, and can outperform the baselines for adversarial attacks with the combination of projected gradient descent (PGD).

关键词：

来源：评论

学校读者我要写书评

暂无评论

Buffered asynchronous SGD for byzantine learning

The Journal of Machine Learning Research

引用

The Journal of Machine Learning Research 2023年第1期24卷 9761-9822页

作者： Yi-Rui Yang Wu-Jun Li National Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University Nanjing China

Distributed learning has become a hot research topic due to its wide application in cluster-based large-scale learning, federated learning, edge computing, and so on. Most traditional distributed learning methods typically assume no failure or attack. However, many unexpected cases, such as communication failure and even malicious attack, may happen in real applications. Hence, Byzantine learning (BL), which refers to distributed learning with failure or attack, has recently attracted much attention. Most existing BL methods are synchronous, which are impractical in some applications due to heterogeneous or offline workers. In these cases, asynchronous BL (ABL) is usually preferred. In this paper, we propose a novel method, called buffered asynchronous stochastic gradient descent (BASGD), for ABL. To the best of our knowledge, BASGD is the first ABL method that can resist non-omniscient attacks without storing any instances on the server. Furthermore, we also propose an improved variant of BASGD, called BASGD with momentum (BASGDm), by introducing local momentum into BASGD. Compared with those methods which need to store instances on server, BASGD and BASGDm have a wider scope of application. Both BASGD and BASGDm are compatible with various aggregation rules. Moreover, both BASGD and BASGDm are proved to be convergent and able to resist failure or attack. Empirical results show that our methods significantly outperform existing ABL baselines when there exists failure or attack on workers.

关键词： distributed machine learning momentum asynchronous Byzantine learning buffer stochastic gradient descent

来源：评论

学校读者我要写书评

暂无评论

InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learning

arXiv

引用

arXiv 2024年

作者： Liang, Yan-Shuo Li, Wu-Jun National Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University China

Continual learning requires the model to learn multiple tasks sequentially. In continual learning, the model should possess the ability to maintain its performance on old tasks (stability) and the ability to adapt to new tasks continuously (plasticity). Recently, parameter-efficient fine-tuning (PEFT), which involves freezing a pre-trained model and injecting a small number of learnable parameters to adapt to downstream tasks, has gained increasing popularity in continual learning. Although existing continual learning methods based on PEFT have demonstrated superior performance compared to those not based on PEFT, most of them do not consider how to eliminate the interference of the new task on the old tasks, which inhibits the model from making a good trade-off between stability and plasticity. In this work, we propose a new PEFT method, called interference-free low-rank adaptation (InfLoRA), for continual learning. InfLoRA injects a small number of parameters to reparameterize the pre-trained weights and shows that fine-tuning these injected parameters is equivalent to fine-tuning the pre-trained weights within a subspace. Furthermore, InfLoRA designs this subspace to eliminate the interference of the new task on the old tasks, making a good trade-off between stability and plasticity. Experimental results show that InfLoRA outperforms existing state-of-the-art continual learning methods on multiple datasets. Code is available at https://***/liangyanshuo/InfLoRA. Copyright © 2024, The Authors. All rights reserved.

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

Energy-Guided Diffusion Sampling for Offline-to-Online Reinforcement Learning 41

Energy-Guided Diffusion Sampling for Offline-to-Online Reinf...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Liu, Xu-Hui Liu, Tian-Shuo Jiang, Shengyi Chen, Ruifeng Zhang, Zhilong Chen, Xinwei Yu, Yang National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China Polixir Technologies China Department of Computer Science The University of Hong Kong Hong Kong

Combining offline and online reinforcement learning (RL) techniques is indeed crucial for achieving efficient and safe learning where data acquisition is expensive. Existing methods replay offline data directly in the online phase, resulting in a significant challenge of data distribution shift and subsequently causing inefficiency in online fine-tuning. To address this issue, we introduce an innovative approach, Energy-guided DIffusion Sampling (EDIS), which utilizes a diffusion model to extract prior knowledge from the offline dataset and employs energy functions to distill this knowledge for enhanced data generation in the online phase. The theoretical analysis demonstrates that EDIS exhibits reduced suboptimality compared to solely utilizing online data or directly reusing offline data. EDIS is a plug-in approach and can be combined with existing methods in offline-to-online RL setting. By implementing EDIS to off-the-shelf methods Cal-QL and IQL, we observe a notable 20% average improvement in empirical performance on MuJoCo, AntMaze, and Adroit environments. Code is available at https://***/liuxhym/EDIS. Copyright 2024 by the author(s)

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Differentiate Xp11.2 Translocation Renal Cell Carcinoma from Computed Tomography Images and Clinical Data with ResNet-18 CNN and XGBoost

引用

computer Modeling in Engineering & sciences 2023年第7期136卷 347-362页

作者： Yanwen Lu Wenliang Ma Xiang Dong Mackenzie Brown Tong Lu Weidong Gan Department of Urology Nanjing Drum Tower HospitalThe Affiliated Hospital of Nanjing University Medical SchoolNanjing210008China School of Data Science Perdana UniversitySerdang43400Malaysia State Key Laboratory for Novel Software Technology Nanjing UniversityNanjing210008China

This study aims to apply ResNet-18 convolutional neural network(CNN)and XGBoost to preoperative computed tomography(CT)images and clinical data for distinguishing Xp11.2 translocation renal cell carcinoma(Xp11.2 tRCC)from common subtypes of renal cell carcinoma(RCC)in order to provide patients with individualized treatment *** from45 patients with Xp11.2 tRCC fromJanuary 2007 to December 2021 are *** cell RCC(ccRCC),papillary RCC(pRCC),or chromophobe RCC(chRCC)can be detected from each *** images are acquired in the following three phases:unenhanced,corticomedullary,and nephrographic.A unified framework is proposed for the classification of renal *** this framework,ResNet-18 CNN is employed to classify renal cancers with CT images,while XGBoost is adopted with clinical *** demonstrate that,if applying ResNet-18 CNN or XGBoost singly,the latter outperforms the former,while the framework integrating both technologies performs similarly or better than ***,the possibility of misclassifying Xp11.2 tRCC,pRCC,and chRCC as ccRCC by the proposed framework is much lower than urologists.

关键词： ResNet-18 CNN XGBoost computed tomography TFE3 renal cell carcinoma

来源：评论

学校读者我要写书评

暂无评论

Buffered Asynchronous Secure Aggregation for Cross-Device Federated Learning

arXiv

引用

arXiv 2024年

作者： Wang, Kun Yang, Yi-Rui Li, Wu-Jun National Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University China

Asynchronous federated learning (AFL) is an effective method to address the challenge of device heterogeneity in cross-device federated learning. However, AFL is usually incompatible with existing secure aggregation protocols used to protect user privacy in federated learning because most existing secure aggregation protocols are based on synchronous aggregation. To address this problem, we propose a novel secure aggregation protocol named buffered asynchronous secure aggregation (BASA) in this paper. Compared with existing protocols, BASA is fully compatible with AFL and provides secure aggregation under the condition that each user only needs one round of communication with the server without relying on any synchronous interaction among users. Based on BASA, we propose the first AFL method which achieves secure aggregation without extra requirements on hardware. We empirically demonstrate that BASA outperforms existing secure aggregation protocols for cross-device federated learning in terms of training efficiency and scalability. Copyright © 2024, The Authors. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

A Combinatorial Interaction Testing Method for Multi-Label Image Classifier

A Combinatorial Interaction Testing Method for Multi-Label I...

引用

International Symposium on software Reliability Engineering (ISSRE)

作者： Peng Wang Shengyou Hu Huayao Wu Xintao Niu Changhai Nie Lin Chen State Key Laboratory for Novel Software Technology and School of Computer Science Nanjing University China

ISBN: (数字)9798350353884

ISBN: (纸本)9798350353891

Multi-label image classification is a critical task in computer vision, in which the correlations between labels are typically exploited by modern classifiers for an effective classification. In this study, we propose LV-CIT, a black-box testing method that applies Combinatorial Interaction Testing (CIT) to systematically test the ability of classifiers to handle such correlations. Specifically, LV-CIT views each label of the label space as an input-parameter taking binary values (indicating whether an object appears in an image), and manages to generate a label value covering array as the set of test cases to cover certain combinations of label values. Then, for each test case, LV-CIT relies on an object library to generate composite test images that perfectly match the specified labels, and reports classification errors if such labels cannot be correctly recognised. The experimental results on two popular datasets with six state-of-the-art image classifiers show that LV-CIT is more efficient than the existing CIT tools in generating label value covering arrays. LV-CIT is also effective in errors revelation, as it can find 111% more errors by using 20% fewer test images than the existing methods for testing multi-label image classifiers.

关键词： Correlation Image recognition Source coding Neural networks Closed box Libraries Reproducibility of results software reliability Testing Image classification

来源：评论

学校读者我要写书评

暂无评论

Color Image Steganalysis Algorithm Based on Quaternion Convolutional Neural Network 30th

Color Image Steganalysis Algorithm Based on Quaternion Convo...

引用

30th International Conference on Computational and Experimental Engineering and sciences, ICCES 2024

作者： Xu, Meng Wang, Jinwei State Key Laboratory of Mathematical Engineering and Advanced Computing Henan450001 China Department of Computer and Software Nanjing University of Information Science and Technology Nanjing210044 China

ISBN: (纸本)9783031687747

Most of the images on the Internet are color images, and steganalysis of color images is a very critical issue in the field of steganalysis. The current proposed color image steganalysis features mainly rely on manual design, and the steganalysis features do not fully consider the internal connections between the three channels of color images. Most steganalyzers based on deep learning are also designed for grayscale images. In recent years, the continuous introduction of advanced steganography techniques for color images has also brought more severe challenges to the steganography analysis of color images. Quaternions are a tool that can effectively represent color images, and utilizing quaternion transformations can fully exploit the correlation between color image channels. Combining quaternion theory with convolutional neural networks, this paper proposed a color image steganalysis algorithm based on quaternion convolutional neural networks. Firstly, the image is represented by quaternion, and then input into the quaternion convolutional neural network. The convolution kernel in the network is composed of quaternions, and then perform convolution and pooling. The experimental results have demonstrated the significant effectiveness of the proposed network. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Task assignment for social-oriented crowdsourcing

引用

Frontiers of computer science 2021年第2期15卷 39-49页

作者： Gang WU Zhiyong CHEN Jia LIU Donghong HAN Baiyou QIAO School of Computer Science and Engineering Northeastern UniversityShenyang110004China State Key Laboratory for Novel Software Technology Nanjing UniversityNanjing210023China

Crowdsourcing has become an efficient measure to solve machine-hard problems by embracing group wisdom,in which tasks are disseminated and assigned to a group of workers in the way of open *** social relationships formed during this process may in turn contribute to the completion of future *** this sense,it is necessary to take social factors into consideration in the research of ***,there is little work on the interactions between social relationships and crowdsourcing *** this paper,we propose to study such interactions in those social-oriented crowdsourcing systems from the perspective of task assignment.A prototype system is built to help users publish,assign,accept,and accomplish location-based crowdsourcing tasks as well as promoting the development and utilization of social relationships during the ***,in order to exploit the potential relationships between crowdsourcing workers and tasks,we propose a“worker-task”accuracy estimation algorithm based on a graph model that joints the factorized matrixes of both the user social networks and the history“worker-task”*** the worker-task accuracy estimation matrix,a group of optimal worker candidates is efficiently chosen for a task,and a greedy task assignment algorithm is proposed to further the matching of worker-task pairs among multiple crowdsourcing tasks so as to maximize the overall *** with the similarity based task assignment algorithm,experimental results show that the average recommendation success rate increased by 3.67%;the average task completion rate increased by 6.17%;the number of new friends added per week increased from 7.4 to 10.5;and the average task acceptance time decreased by 8.5 seconds.

关键词： crowdsourcing social networks task assignment

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Machine Unlearning 1

引用

17th International Conference on Learning and Intelligent Optimization, LION-17 2023

作者： Zhu, HongBin Xia, YuXiao Li, YunZhao Li, Wei Liu, Kang Gao, Xianzhou Big Data Center State Grid of China Beijing100012 China State Grid Smart Grid Research Institute Shanghai200016 China State Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University Nanjing200016 China

ISBN: (数字)9783031445057

ISBN: (纸本)9783031445040

In recent years, deep neural networks have enjoyed tremendous success in industry and academia, especially for their applications in visual recognition and natural language processing. While large-scale deep models bring incredible performance, their massive data requirements pose a huge threat to data privacy protection. With the growing emphasis on data security, the study of data privacy leakage in machine learning, such as machine unlearning, has become increasingly important. There have been many works on machine unlearning, and other research has proposed training several submodels to speed up the retraining process, by dividing the training data into several disjoint fragments. When the impact of a particular data point in the model is to be removed, the model owner simply retrains the sub-model containing this data point. Nevertheless, current learning methods for machine unlearning are still not widely used due to model applicability, usage overhead, etc. Based on this situation, we propose a novel hierarchical learning method, Hierarchical Machine Unlearning (HMU), with the known distribution of unlearning requests. Compared with previous methods, ours has better efficiency. Using the known distribution, the data can be partitioned and sorted, thus reducing the overhead in the data deletion process. We propose to train the model using the hierarchical data set after partitioning, which further reduces the loss of prediction accuracy of the existing methods. It is also combined with incremental learning methods to speed up the training process. Finally, the effectiveness and efficiency of the method proposed in this paper are verified by multiple experiments. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Data privacy

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：