检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Lei, Shiye Zhang, Sen Tao, Dacheng School of Computer Science The University of Sydney Australia College of Computing & Data Science Nanyang Technological University Singapore

Massive reinforcement learning (RL) data are typically collected to train policies offline without the need for interactions, but the large data volume can cause training inefficiencies. To tackle this issue, we formulate offline behavior distillation (OBD), which synthesizes limited expert behavioral data from sub-optimal RL data, enabling rapid policy learning. We propose two naive OBD objectives, DBC and PBC, which measure distillation performance via the decision difference between policies trained on distilled data and either offline data or a near-expert policy. Due to intractable bi-level optimization, the OBD objective is difficult to minimize to small values, which deteriorates PBC by its distillation performance guarantee with quadratic discount complexity O(1/(1 − γ)2). We theoretically establish the equivalence between the policy performance and action-value weighted decision difference, and introduce action-value weighted PBC (Av-PBC) as a more effective OBD objective. By optimizing the weighted decision difference, Av-PBC achieves a superior distillation guarantee with linear discount complexity O(1/(1 − γ)). Extensive experiments on multiple D4RL datasets reveal that Av-PBC offers significant improvements in OBD performance, fast distillation convergence speed, and robust cross-architecture/optimizer generalization. The code is available at https://***/LeavesLei/OBD. © 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

Enhancing Automated COVID Diagnosis from Chest X-rays using Convolutional Neural Networks and Transfer Learning 5

Enhancing Automated COVID Diagnosis from Chest X-rays using ...

引用

5th International Conference on Advancements in Computational sciences, ICACS 2024

作者： Bukhari, Muhammad Abdullah Shah Siddiqui, Muhammad Ahmad Khalique, Sarmad Bukhari, Faisal Iqbal, Waheed University of the Punjab Department of Computer Science Lahore Pakistan University of the Punjab Department of Data Science Lahore Pakistan

ISBN: (纸本)9798350394764

We propose an approach for the early detection of COVID-19 and other related lung diseases using artificial intelligence (AI) and deep learning-based methods. The proposed approach involves utilizing transfer learning over convolutional neural networks (CNNs) for the classification of chest X-ray images as normal or COVID-19 positive. To address the limited availability of X-ray images, we employed data augmentation techniques to expand the dataset size and enhance our model's performance on unseen data. We used variations of four pre-trained models: DenseNet, Inception, ResNet, and VGG, and found that DenseNet169 demonstrated the best performance with a validation accuracy of 100%. © 2024 IEEE.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Peer Collaboration in DBLP Using Graph Convolutional Network

引用

SN computer science 2025年第1期6卷 1-7页

作者： Kumar, Chintoo Dar, Showkat Ahmad Batchu, Charishma Singamaneni, Kundan Sai Panguluri, Venkata Lakshmi Alekhya Department of Artificial Intelligence and Data Science GITAM Karnataka Bengaluru 562163 India Department of Computer Science and Engineering GITAM Karnataka Bengaluru 562163 India

The increasing popularity of Graph-based neural network architectures plays a pivotal role in providing promising results in applications, viz., Friendship networks, Co-authorship networks, Product recommendations, etc. Many machine learning tasks can be addressed when dealing with graph representation, i.e., graph classification, node classification, link prediction, graph embedding, etc. The traditional rule-based methods have many drawbacks in comparison to graph representation learning, such as it requires manual feature engineering and struggles with unseen cases (generalizability), etc., in the case of rule-based methods. At the same time, graph representation learning overcomes those issues profoundly. The advent of graph-based deep neural network architectures inspired us to complete a graph classification problem to explore clusters of researchers who worked together in academic publications. We built the model by using a graph convolutional network to solve a multi-classification problem. We conducted a comparative study to check the efficacy of graph representation learning by GCNs on the DBLP dataset. A comparative study shows that the built model outperforms the baseline models. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： DBLP Graph convolutional network Graph mining Social networks

来源：评论

学校读者我要写书评

暂无评论

Lay CO:Achieving Least Lossy Accuracy for Most Efficient RRAM-Based Deep Neural Network Accelerator via Layer-Centric Co-Optimization

引用

计算机科学技术学报（英文版） 2023年第2期38卷 328-347页

作者： Shao-Feng Zhao Fang Wang Bo Liu Dan Feng Yang Liu Wuhan National Laboratory for Optoelectronics Huazhong University of Science and TechnologyWuhan 430074China School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Cloud Computing and Big Data Institute Henan University of Economics and LawZhengzhou 450001China Wuhan National Laboratory for Optoelectronics Huazhong University of Science and TechnologyWuhan 430074China School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Research Institute of Huazhong University of Science and Technology in Shenzhen Shenzhen 518057China School of Computer and Artificial Intelligence Zhengzhou UniversityZhengzhou 450001China Wuhan National Laboratory for Optoelectronics Huazhong University of Science and TechnologyWuhan 430074China School of Computer Science and Technology Huazhong University of Science and TechnologyWuhan 430074China Cloud Computing and Big Data Institute Henan University of Economics and LawZhengzhou 450001China

Resistive random access memory(RRAM)enables the functionality of operating massively parallel dot prod-ucts and ***-based accelerator is such an effective approach to bridging the gap between Internet of Things devices'constrained resources and deep neural networks'tremendous *** to the huge overhead of Analog to Digital(A/D)and digital accumulations,analog RRAM buffer is introduced to extend the processing in analog and in *** analog RRAM buffer offers potential solutions to A/D conversion issues,the energy consumption is still challenging in resource-constrained environments,especially with enormous intermediate data ***,criti-cal concerns over endurance must also be resolved before the RRAM buffer could be frequently used in reality for DNN in-ference *** we propose LayCO,a layer-centric co-optimizing scheme to address the energy and endurance con-cerns altogether while strictly providing an inference accuracy *** relies on two key ideas:1)co-optimizing with reduced supply voltage and reduced bit-width of accelerator architectures to increase the DNN's error tolerance and achieve the accelerator's energy efficiency,and 2)efficiently mapping and swapping individual DNN data to a correspond-ing RRAM partition in a way that meets the endurance *** evaluation with representative DNN models demonstrates that LayCO outperforms the baseline RRAM buffer based accelerator by 27x improvement in energy effi-ciency(over TIMELY-like configuration),308x in lifetime prolongation and 6x in area reduction(over RAQ)while main-taining the DNN accuracy loss less than 1％.

关键词： deep learning accelerator resistive random access memory (RRAM) energy efficiency error tolerance

来源：评论

学校读者我要写书评

暂无评论

Enhance Transferability of Adversarial Examples with Model Architecture 48

Enhance Transferability of Adversarial Examples with Model A...

引用

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Fan, Mingyuan Guo, Wenzhong Ying, Zuobin Liu, Ximeng Fuzhou University College of Computer and Data Science China City University of Macau Faculty of Data Science China

ISBN: (纸本)9781728163277

Transferability of adversarial examples is of critical importance to launch black-box adversarial attacks, where attackers are only allowed to access the output of the target model. However, under such a challenging but practical setting, the crafted adversarial examples are always prone to overfitting to the proxy model employed, presenting poor transferability. In this paper, we suggest alleviating the overfitting issue from a novel perspective, i.e., designing a fitted model architecture. Specifically, delving the bottom of the cause of poor transferability, we arguably decompose and reconstruct the existing model architecture into an effective model architecture, namely multi-track model architecture (MMA). The adversarial examples crafted on the MMA can maximumly relieve the effect of model-specified features to it and toward the vulnerable directions adopted by diverse architectures. Extensive experimental evaluation demonstrates that the transferability of adversarial examples based on the MMA significantly surpass other state-of-the-art model architectures by up to 40% with comparable overhead. © 2023 IEEE.

关键词： Architecture

来源：评论

学校读者我要写书评

暂无评论

Dynamic Backup Sharing Scheme of Service Function Chains in NFV

引用

China Communications 2022年第5期19卷 178-190页

作者： Dong Zhang Zhifan Zheng Xiang Lin Xiang Chen Chunming Wu College of Computer and Data Science Fuzhou UniversityFuzhou 350108China College of Computer Science and Technology Zhejiang UniversityHangzhou 310058China

Service function chains(SFC)mapping takes the responsibility for managing virtual network functions(VNFs).In SFC mapping,existing solutions duplicate VNFs with redundant instances to provide high availability in response to ***,as a compromise,these solutions result in high resource consumption due to device *** this paper,we propose a novel method named dynamic backup sharing(DBS)that allows SFCs to dynamically share backups to reduce resource *** formulates the problem of sharing backups among different VNFs as an integer linear programming(ILP).Thereafter,we design a novel online algorithm based on dynamic programming to solve the *** experimental results indicate that DBS outperforms state-ofthe-art works by reducing resource consumption and improving the number of accepted requests.

关键词： network function virtualization service function chain mapping reliability backup sharing

来源：评论

学校读者我要写书评

暂无评论

Deep learning-based method for detection of copy-move forgery in videos

引用

Neural Computing and Applications 2025年第14期37卷 8451-8464页

作者： Elbarougy, Reda Abdelfatah, Osama Behery, G.M. El-Badry, Noha M. Department of Artificial Intelligence and Data Science College of Computer Science and Engineering University of Ha’il Ha’il81481 Saudi Arabia Department of Information Technology Faculty of Computer and Artificial Intelligence Damietta University New Damietta Kafr Saad Egypt Department of Computer Science Faculty of Computer and Artificial Intelligence Damietta University New Damietta Kafr Saad Egypt Department of Mathematics Faculty of Science Damietta University New Damietta Kafr Saad Egypt

Video forgery is one of the most serious problems affecting the credibility and reliability of video content. Therefore, detecting video forgery presents a major challenge for researchers due to the diversity of forgery types, the modernity of the programs used in forgery operations, and the abundance of information and content present in videos. The seriousness of this issue arises from the widespread use of videos in vital fields that require very high accuracy with no room for doubt or error, such as courtrooms, journalism, and others. Copy-move forgery is one of the most common and dangerous types of video forgery because of the difficulty in identifying it with the naked eye and the great diversity of forgery techniques involved in this particular type. One of the biggest challenges researchers have faced in the past is the complexity of the steps required to detect forgery in videos, leading to significant computational complexity and time consumption. The proposed method also aims to achieve better results than previous methods while reducing computational operations. Ultimately, forgery is detected with great efficiency. Compared to previous methods used for detecting copy-move forgery in the Rewind dataset, the proposed method achieves the highest F1, reaching 0.86, with a significant difference of 0.13 compared to the best result of the previous methods. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

RLHR: A Framework for Driving Dynamically Adaptable Questionnaires and Profiling People Using Reinforcement Learning

RLHR: A Framework for Driving Dynamically Adaptable Question...

引用

19th International Conference on Software Technologies, ICSOFT 2024

作者： Paduraru, Ciprian Patilea, Catalina Camelia Stefanescu, Alin Department of Computer Science University of Bucharest Romania Institute for Logic and Data Science Romania

ISBN: (纸本)9789897587061

In today’s corporate landscape, the creation of questionnaires, surveys or evaluation forms for employees is a widespread practice. These tools are regularly used to check various aspects such as motivation, opportunities for improvement, satisfaction levels and even potential cybersecurity risks. A common limitation lies in their generic nature: they often lack personalization and rely on predetermined questions. Our research focuses on improving this process by introducing AI agents based on reinforcement learning. These agents dynamically adapt the content of surveys to each person based on their unique personality traits. Our framework is open source and can be seamlessly integrated into various use cases in different industries or academic research. To evaluate the effectiveness of the approach, we tackle a real-life scenario: the detection of potentially inappropriate behavior in the workplace. In this context, the reinforcement learning-based AI agents function like human recruiters and create personalized surveys. The results are encouraging, as they show that our decision algorithms for content selection are very similar to those of recruiters. The open-source framework also includes tools for detailed post-analysis for further decision making and explanation of the results. © 2024 by SCITEPRESS – science and Technology Publications, Lda.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

ARCHER:a ReRAM-based accelerator for compressed recommendation systems

引用

Frontiers of computer science 2024年第5期18卷 147-160页

作者： Xinyang SHEN Xiaofei LIAO Long ZHENG Yu HUANG Dan CHEN Hai JIN National Engineering Research Center for Big Data Technology and System Services Computing Technology and System LabClusters and Grid Computing LabSchool of Computer Science and TechnologyHuazhong University of Science and TechnologyWuhan 430074China

Modern recommendation systems are widely used in modern data *** random and sparse embedding lookup operations are the main performance bottleneck for processing recommendation systems on traditional platforms as they induce abundant data movements between computing units and ***-based processing-in-memory(PIM)can resolve this problem by processing embedding vectors where they are ***,the embedding table can easily exceed the capacity limit of a monolithic ReRAM-based PIM chip,which induces off-chip accesses that may offset the PIM ***,we deploy the decomposed model on-chip and leverage the high computing efficiency of ReRAM to compensate for the decompression performance *** this paper,we propose ARCHER,a ReRAM-based PIM architecture that implements fully yon-chip recommendations under resource ***,we make a full analysis of the computation pattern and access pattern on the decomposed *** on the computation pattern,we unify the operations of each layer of the decomposed model in multiply-and-accumulate *** on the access observation,we propose a hierarchical mapping schema and a specialized hardware design to maximize resource *** the unified computation and mapping strategy,we can coordinatethe inter-processing elements *** evaluation shows that ARCHER outperforms the state-of-the-art GPU-based DLRM system,the state-of-the-art near-memory processing recommendation system RecNMP,and the ReRAM-based recommendation accelerator REREC by 15.79×,2.21×,and 1.21× in terms of performance and 56.06×,6.45×,and 1.71× in terms of energy savings,respectively.

关键词： recommendation system ReRAM processing-in-memory embedding layer

来源：评论

学校读者我要写书评

暂无评论

Joint Internal Multi-Interest Exploration and External Domain Alignment for Cross Domain Sequential Recommendation 23

Joint Internal Multi-Interest Exploration and External Domai...

引用

32nd ACM World Wide Web Conference, WWW 2023

作者： Liu, Weiming Zheng, Xiaolin Chen, Chaochao Su, Jiajie Liao, Xinting Hu, Mengling Tan, Yanchao College of Computer Science Zhejiang University China College of Computer and Data Science Fuzhou University China

ISBN: (纸本)9781450394161

Sequential Cross-Domain Recommendation (CDR) has been popularly studied to utilize different domain knowledge and users' historical behaviors for the next-item prediction. In this paper, we focus on the cross-domain sequential recommendation problem. This commonly exist problem is rather challenging from two perspectives, i.e., the implicit user historical rating sequences are difficult in modeling and the users/items on different domains are mostly non-overlapped. Most previous sequential CDR approaches cannot solve the cross-domain sequential recommendation problem well, since (1) they cannot sufficiently depict the users' actual preferences, (2) they cannot leverage and transfer useful knowledge across domains. To tackle the above issues, we propose joint Internal multi-interest exploration and External domain alignment for cross domain Sequential Recommendation model (IESRec). IESRec includes two main modules, i.e., internal multi-interest exploration module and external domain alignment module. To reflect the users' diverse characteristics with multi-interests evolution, we first propose internal temporal optimal transport method in the internal multi-interest exploration module. We further propose external alignment optimal transport method in the external domain alignment module to reduce domain discrepancy for the item embeddings. Our empirical studies on Amazon datasets demonstrate that IESRec significantly outperforms the state-of-the-art models. © 2023 ACM.

关键词： Alignment

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：