检索结果-内蒙古大学图书馆

5th International Conference on Artificial intelligence and Computer engineering, ICAICE 2024

作者： Zhang, Chuancai Qu, Dan Du, Liming Yang, Kaiyuan School of Information Systems Engineering University of Information Engineering Zhengzhou China Laboratory for Advanced Computing and Intelligence Engineering Zhengzhou China School of Cyberspace Security Zhengzhou University Zhengzhou China

ISBN: (纸本)9798331528911

In Unsupervised Neural Machine Translation (UNMT) tasks, the lack of extensive parallel corpora makes it challenging for the model to directly optimize the correspondence between the source and target languages. UNMT models primarily rely on learning from monolingual data, making it difficult to fully capture cross-linguistic semantic alignment and consistency during training. Therefore, effectively enhancing semantic alignment between the source and target languages under unsupervised conditions has become a key challenge in the field of UNMT. To address this issue, this study proposes a semantic alignment and consistency enhancement mechanism based on contrastive learning, aiming to improve the model's ability to capture cross-lingual semantic differences by constructing an optimization objective focused on semantic consistency. The method utilizes a pretrained cross-lingual model to generate semantic embeddings for both the source and target languages, then designs a contrastive learning-based loss function that minimizes the semantic distance between correct translations and source sentences while increasing the semantic distance between incorrect translations and source sentences, thereby enhancing the model's performance in semantic consistency. The experiments validate the effectiveness and broad applicability of the semantic consistency optimization method based on contrastive learning, demonstrating great potential and advantages, especially in multilingual tasks of UNMT. The model achieved significant improvements in BLEU scores across various language pairs. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Software Diversification Protection Methods for Binary Programs

Software Diversification Protection Methods for Binary Progr...

引用

6th International Conference on Intelligent computing and Signal Processing (ICSP)

作者： Yingchao Chen Junchao Wang Xin Zhou Jianmin Pang Laboratory for Advanced Computing and Intelligence Engineering Zhengzhou China

ISBN: (数字)9798350376548

ISBN: (纸本)9798350376555

Software diversification is an effective software protection method against reverse engineering and code reuse attacks, which can provide heterogeneous redundant execution bodies for mimetic defense mechanisms. Most existing software diversification methods require access to the source code, which can provide defenders with more valuable information for devising effective defense strategies. However, due to commercial copyright and the purpose of preventing software piracy, developers often keep their software closed-source, making it difficult to access the source code. Therefore, a method called R2BF (ReCooking and Randomizing Binary File) is proposed, which combines decompilation and diversification compilation techniques to address the difficulty of obtaining the source code. This method involves diversifying the source code through compilation to achieve software diversification for binary executable files. Security and performance testing of diversified C programs has shown that diversified binary executable files can mitigate vulnerabilities resulting from code reuse, validating that this method can enhance the security of binary programs and is feasible for software protection. The diversified binary executable files are nearly identical to the original, non-diversified files in terms of file size and execution time, and may even exhibit acceleration and optimization effects in certain scenarios.

关键词： Software testing Codes Source coding Software protection Software algorithms Reverse engineering Signal processing algorithms Software Security Optimization

来源：评论

学校读者我要写书评

暂无评论

Continual learning with Bayesian model based on a fixed pre-trained feature extractor

引用

Visual intelligence 2023年第1期1卷 1-14页

作者： Yang, Yang Cui, Zhiying Xu, Junjie Zhong, Changhong Zheng, Wei-Shi Wang, Ruixuan School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Department of Network Intelligence Peng Cheng Laboratory Shenzhen China Key Laboratory of Machine Intelligence and Advanced Computing MOE Guangzhou China

Deep learning has shown its human-level performance in various applications. However, current deep learning models are characterized by catastrophic forgetting of old knowledge when learning new classes. This poses a challenge such as in intelligent diagnosis systems where initially only training data of a limited number of diseases are available. In this case, updating the intelligent system with data of new diseases would inevitably downgrade its performance on previously learned diseases. Inspired by the process of learning new knowledge in human brains, we propose a Bayesian generative model for continual learning built on a fixed pre-trained feature extractor. In this model, knowledge of each old class can be compactly represented by a collection of statistical distributions, e.g., with Gaussian mixture models, and naturally kept from forgetting in continual learning over time. Unlike existing class-incremental learning methods, the proposed approach is not sensitive to the continual learning process and can be additionally well applied to the data-incremental learning scenario. Experiments on multiple medical and natural image classification tasks reveal that the proposed approach outperforms state-of-the-art approaches that even keep some images of old classes during continual learning of new classes. © The Author(s) 2023.

关键词： Bayesian model Continual learning Fixed feature extractor Generative approach

来源：评论

学校读者我要写书评

暂无评论

TexCIL: Text-Guided Continual Learning of Disease with Vision-Language Model

TexCIL: Text-Guided Continual Learning of Disease with Visio...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Zhang, Wentao Zhao, Defeng Zheng, Wei-Shi Wang, Ruixuan Sun Yat-sen University School of Computer Science and Engineering Guangzhou China Peng Cheng Laboratory Shenzhen China Moe Key Laboratory of Machine Intelligence and Advanced Computing Guangzhou China

ISBN: (纸本)9798350386226

Current intelligent diagnostic systems often catastrophically forget old knowledge when learning new diseases only from the training dataset of the new diseases. Inspired by human learning of visual classes with the effective help of language, we propose a continual learning framework based on a pre-trained visual-language model (VLM) without storing any image of previously learned diseases. In this framework, textual prior knowledge of each new disease can be obtained by utilizing the frozen VLM's text encoder, and then used to guide the visual learning of the new disease. This framework innovatively utilizes the textual prior knowledge of all previously learned diseases as out-of-distribution (OOD) information to help differentiate currently being-learned diseases from others. Extensive empirical evaluations on both medical and natural image datasets confirm the superiority of the proposed method over existing state-of-the-art methods in continual learning of new visual classes. The source code is available at https://***/OpenMedIA/TexCIL. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Path test data generation using adaptive simulated annealing particle swarm optimization

引用

Soft computing 2024年第17-18期28卷 9587-9607页

作者： Jiao, Chongyang Zhou, Qinglei State Key Laboratory of Mathematical Engineering and Advanced Computing PLA Strategic Support Force Information Engineering University Zhengzhou450001 China Henan Information Engineering School Zhengzhou Vocational College of Industrial Safety Zhengzhou450000 China School of Computer and Artificial Intelligence Zhengzhou University Zhengzhou450001 China

Software testing is an effective means of ensuring software quality. The cost of software testing is the main component of the total cost of software development. The generation of test data is very important in testing, because the efficiency of testing depends on the test data used. A significant part of software testing automation is the test data generation. How to automatically generate test datasets is still an open problem. Search-based software testing (SBST) is a process of generating test data that employs meta-heuristic methods to resolve tough NP-hard problems. The test data generation problem belongs to NP-hard problems. This paper focuses on the automatic generation of test data for path coverage based on control flow criteria in structural testing. Some meta-heuristic algorithms have been used to search for software test data. To advance the efficiency and effectiveness of path test data generation, an adaptive simulated annealing particle swarm optimization (ASAPSO) algorithm is proposed. The probabilistic jumping property of simulated annealing (SA) algorithm is introduced into the particle swarm optimization (PSO) algorithm to make the particle accept bad solution at a certain probability during the searching process, and therefore boost the capacity of the proposed algorithm to bounce from the local optimum. The fitness function is constructed by the branch function superposition to properly guide the search process. In addition, to improve the convergence speed of the algorithm, an adaptive adjustment scheme based on inertial weight and an adjustment scheme based on learning factor are proposed as well. The experimental results show that the proposed method can effectually avoid premature convergence and effectively upgrades the efficiency of generating test data automatically. It is also competitive in solving other complicated optimization problems. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Natur

关键词： Software testing

来源：评论

学校读者我要写书评

暂无评论

EB-SNN: An Ensemble Binary Spiking Neural Network for Visual Recognition 27th

EB-SNN: An Ensemble Binary Spiking Neural Network for Visua...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Li, Xinjie Tang, Jianxiong Lai, Jianhuang School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Guangzhou510555 China Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education Guangzhou China

ISBN: (纸本)9783031781858

In recent years, spiking neural networks (SNNs) have gained significant attention in visual recognition tasks due to the low computational energy. However, most SNNs have a large number of parameters, which limits their use on resource-limited devices. In this paper, we propose an Ensemble Binary Spiking Neural Network (EB-SNN) for accurate and memory-friendly visual recognition. The EB-SNN is modeled by Ensemble Binary Weights (EBW) module, which integrates multiple binary weights for lightweight SNN modeling. Meanwhile, we propose Knowledge Alignment Strategy to ensure that the EB-SNN can approximate a well-trained SNN for good performance. Experimental results show that the EB-SNN can achieve accuracy of 95.39% on CIFAR10, using 9.3% memory of full-precision SNN. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Local binary pattern

来源：评论

学校读者我要写书评

暂无评论

Breaking Information Silos: Global Guided Task Prediction for Class-Incremental Learning 27th

Breaking Information Silos: Global Guided Task Prediction fo...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Hu, Chaoshun Ye, Biaohua Chen, Zixuan Lai, Jian-Huang School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Guangzhou510555 China Ministry of Education Key Laboratory of Machine Intelligence and Advanced Computing Guangzhou China

ISBN: (纸本)9783031781889

Class-incremental learning (CIL) aims to learn a series of tasks sequentially, each introducing several new categories. Because providing the task labels during inference can significantly increase accuracy, many approaches attempt to predict task labels in CIL. However, existing works focus on learning local information and overlook the importance of global information. The absence of global guidance leads to the formation of information silos across disparate tasks, resulting in potential inter-task interference. To break information silos, we propose a method called Global Guided Task Prediction (GGTP) to introduce global information. Our method consists of two modules. The local de-redundant module aims to reduce information redundancy across different tasks from a global perspective. We combine dual encoders and feature decorrelation loss to effectively reduce redundancy while minimizing catastrophic forgetting. The global information module explicitly extracts global information to serve as auxiliary information guiding task prediction. We extract the most important information from all local information through a global encoder and then aggregate them to form global information. Extensive experiments validate the effectiveness of our method, which achieves state-of-the-art results. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Redundancy

来源：评论

学校读者我要写书评

暂无评论

Grasp as You Say: Language-guided Dexterous Grasp Generation 38

Grasp as You Say: Language-guided Dexterous Grasp Generation

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Wei, Yi-Lin Jiang, Jian-Jian Xing, Chengyi Tan, Xian-Tuo Wu, Xiao-Ming Li, Hao Cutkosky, Mark Zheng, Wei-Shi School of Computer Science and Engineering Sun Yat-sen University China Stanford University United States Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education China

This paper explores a novel task "Dexterous Grasp as You Say" (DexGYS), enabling robots to perform dexterous grasping based on human commands expressed in natural language. However, the development of this field is hindered by the lack of datasets with natural human guidance;thus, we propose a language-guided dexterous grasp dataset, named DexGYSNet, offering high-quality dexterous grasp annotations along with flexible and fine-grained human language guidance. Our dataset construction is cost-efficient, with the carefully-design hand-object interaction retargeting strategy, and the LLM-assisted language guidance annotation system. Equipped with this dataset, we introduce the DexGYSGrasp framework for generating dexterous grasps based on human language instructions, with the capability of producing grasps that are intent-aligned, high quality and diversity. To achieve this capability, our framework decomposes the complex learning process into two manageable progressive objectives and introduce two components to realize them. The first component learns the grasp distribution focusing on intention alignment and generation diversity. And the second component refines the grasp quality while maintaining intention consistency. Extensive experiments are conducted on DexGYSNet and real world environments for validation. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Class-specific Prompts in Vision Transformer for Continual Learning of New Diseases

Class-specific Prompts in Vision Transformer for Continual L...

引用

2023 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2023

作者： Zhao, Defeng Ye, Zejun Zheng, Wei-Shi Wang, Ruixuan Sun Yat-sen University School of Computer Science and Engineering Guangzhou China Peng Cheng Laboratory Department of Network Intelligence Shenzhen China Moe Key Laboratory of Machine Intelligence and Advanced Computing China

ISBN: (纸本)9798350337488

Current intelligent diagnosis systems are often trained to diagnose a small number of diseases and lack the ability of continually learning new disease knowledge. To have such continual learning ability, the deployed intelligent system needs to be continually updated based on training data of only new diseases, without accessing old data of previously learned diseases due to privacy concerns and challenges in data sharing across medical centers. In this study, a novel and effective prompt learning strategy is proposed to help a pretrained and fixed Vision Transformer (ViT) continually learn new diseases. In particular, a set of unique prompts for each disease are effectively learned such that discriminative disease features can be well extracted from the fixed ViT under the instructions of prompts during feature extraction, even though the ViT feature extractor is pretrained in the natural image domain. Extensive empirical evaluations on two medical image datasets and one natural image dataset demonstrate the superior performance of the proposed method. The source code is available at https://***/zhaoedf/CSPrompt. © 2023 IEEE.

关键词： Continual learning Disease diagnosis Prompt learning

来源：评论

学校读者我要写书评

暂无评论

Imputing DNA Methylation by Transferred Learning Based Neural Network

引用

Journal of Computer Science & Technology 2022年第2期37卷 320-329页

作者： Xin-Feng Wang Xiang Zhou Jia-Hua Rao Zhu-Jin Zhang Yue-Dong Yang School of Computer Science and Engineering Sun Yat-sen UniversityGuangzhou 510000China Key Laboratory of Machine Intelligence and Advanced Computing of Ministry of Education(Sun Yat-sen University)Guangzhou 510000 China

DNA methylation is one important epigenetic type to play a vital role in many diseases including *** the development of the high-throughput sequencing technology,there is much progress to disclose the relations of DNA methylation with ***,the analyses of DNA methylation data are challenging due to the missing values caused by the limitations of current *** many methods have been developed to impute the missing values,these methods are mostly based on the correlations between individual samples,and thus are limited for the abnormal samples in *** this study,we present a novel transfer learning based neural network to impute missing DNA methylation data,namely the TDimpute-DNAmeth *** method learns common relations between DNA methylation from pan-cancer samples,and then fine-tunes the learned relations over each specific cancer type for imputing the missing *** on 16 cancer datasets,our method was shown to outperform other commonly-used *** analyses indicated that DNA methylation is related to cancer survival and thus can be used as a biomarker of cancer prognosis.

关键词： neural network transfer learning DNA methylation data imputation survival analysis

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：