检索结果-内蒙古大学图书馆

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Minghao Chen Yihang Li Yanting Yang Shiyu Yu Binbin Lin Xiaofei He School of Computer Science Hangzhou Dianzi University State Key Lab of CAD&CG Zhejiang University School of Software Technology Zhejiang University NingBo Port Group School of Software Technology Zhejiang University and Fullong Inc.

ISBN: (纸本)9798331314385

Large Language Models (LLM) based agents have shown promise in autonomously completing tasks across various domains, e.g., robotics, games, and web navigation. However, these agents typically require elaborate design and expert prompts to solve tasks in specific domains, which limits their adaptability. We introduce AutoManual, a framework enabling LLM agents to autonomously build their understanding through interaction and adapt to new environments. AutoManual categorizes environmental knowledge into diverse rules and optimizes them in an online fashion by two agents: 1) The Planner codes actionable plans based on current rules for interacting with the environment. 2) The Builder updates the rules through a well-structured rule system that facilitates online rule management and essential detail retention. To mitigate hallucinations in managing rules, we introduce a case-conditioned prompting strategy for the Builder. Finally, the Formulator agent compiles these rules into a comprehensive manual. The self-generated manual can not only improve the adaptability but also guide the planning of smaller LLMs while being human-readable. Given only one simple demonstration, AutoManual significantly improves task success rates, achieving 97.4% with GPT-4-turbo and 86.2% with GPT-3.5-turbo on ALFWorld benchmark tasks. The code is available at https://***/minghchen/automanual.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hyperbolic Contrastive Learning with Second Order Sampling for Collaborative Filtering

Hyperbolic Contrastive Learning with Second Order Sampling f...

引用

IEEE International Conference on Web Services (ICWS)

作者： Longchuan Tu Shunmei Meng Lianyong Qi Xiaolong Xu Xuyun Zhang School of Computer Science and Engineering Nanjing University of Science and Technology China State Key Lab. for Novel Software Technology Nanjing University China College of Computer Science and Technology China University of Petroleum(East China) China School of Computer and Software Nanjing University of Information Science and Technology China School of Computing Macquarie University Australia

ISBN: (数字)9798350368550

ISBN: (纸本)9798350368567

Recently, hyperbolic spaces have proven beneficial for service recommendation due to their exponentially growing spatial properties conforming to power-law distributed user-item networks. Among them, the combination of hyperbolic space with graph convolution has achieved great success. However, hyperbolic convolutional models still perform multi-layer convolution in tangent space (Euclidean space), leading to the still inevitable problem of over-smoothing arising from multi-layer convolution. In addition, most of these models randomly draw negative samples from items that users have not interacted with, so that some of the samples obtained may not be well suited for model optimization. To tackle the above challenges, we propose a new Hyperbolic GCN model based on Contrastive Learning and Second-order Reachable sampling for collaborative filtering (HG-CLSR), which improve high quality of representations by exploring the distribution of users and items in hyperbolic space. Specifically, We first introduce a root alignment approach to encourage embeddings to align with the tangent space, thereby reducing distortions during the embedding mapping process in space. Then, we perform contrastive learning in hyperbolic space to motivate the spatial distribution of nodes to better fit the hyperbolic space. Moreover, we sample in the user’s second-order reachable item set, which ensures that the negative sample is more similar to the positive sample, so that the negative node can provide better information for guiding model optimization. Extensive experiments on three real-world datasets demonstrate that the HG-CLSR is significantly superior compared to existing hyperbolic models.

关键词： Degradation Graphical models Convolution Web services Collaborative filtering Design methodology Contrastive learning Distortion Optimization Distribution functions

来源：评论

学校读者我要写书评

暂无评论

ROTATION HAS TWO SIDES: EVALUATING DATA AUGMENTATION FOR DEEP ONE-CLASS CLASSIFICATION 12

ROTATION HAS TWO SIDES: EVALUATING DATA AUGMENTATION FOR DEE...

引用

12th International Conference on Learning Representations, ICLR 2024

作者： Wang, Guodong Wang, Yunhong Bao, Xiuguo Huang, Di State Key Laboratory of Software Development Environment Beihang University Beijing China School of Computer Science and Engineering Beihang University Beijing China Natl. Comp. Net. Emer. Resp. Tech. Team/Coord. Ctr. of China Beijing China

One-class classification (OCC) involves predicting whether a new data is normal or anomalous based solely on the data from a single class during training. Various attempts have been made to learn suitable representations for OCC within a self-supervised framework. Notably, discriminative methods that use geometric visual transformations, such as rotation, to generate pseudo-anomaly samples have exhibited impressive detection performance. Although rotation is commonly viewed as a distribution-shifting transformation and is widely used in the literature, the cause of its effectiveness remains a mystery. In this study, we are the first to make a surprising observation: there exists a strong linear relationship (Pearson's Correlation, r > 0.9) between the accuracy of rotation prediction and the performance of OCC. This suggests that a classifier that effectively distinguishes different rotations is more likely to excel in OCC, and vice versa. The root cause of this phenomenon can be attributed to the transformation bias in the dataset, where representations learned from transformations already present in the dataset tend to be less effective, making it essential to accurately estimate the transformation distribution before utilizing pretext tasks involving these transformations for reliable self-supervised representation learning. To the end, we propose a novel two-stage method to estimate the transformation distribution within the dataset. In the first stage, we learn general representations through standard contrastive pre-training. In the second stage, we select potentially semantics-preserving samples from the entire augmented dataset, which includes all rotations, by employing density matching with the provided reference distribution. By sorting samples based on semantics-preserving versus shifting transformations, we achieve improved performance on OCC benchmarks. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Rotation

来源：评论

学校读者我要写书评

暂无评论

A Privacy-Preserving Decision Tree Evaluation Scheme for Multiple Wearable Devices

A Privacy-Preserving Decision Tree Evaluation Scheme for Mul...

引用

International Conference on computer Supported Cooperative Work in Design

作者： Yuanshun Huang Guihua Duan State Key Laboratory for Novel Software Technology Nanjing University Nanjing China School of Computer Science and Engineering Central South University Changsha China

ISBN: (数字)9798350349184

ISBN: (纸本)9798350349191

Privacy-Preserving Decision Tree (PPDT) evaluation is a typical algorithm in Machine Learning as a Service scenario. For services like health analysis, PPDT evaluation by fusing private data from multiple Wearable Devices is necessary. Homomorphic Encryption (HE) implements PPDT evaluation without revealing locally sensitive data. However, existing HE schemes need better scalability and are incompatible with clients such as Wearable Devices, which have limited computing power, bandwidth, memory, and storage. This paper proposed a noninteractive PPDT evaluation scheme for multiple Wearable Devices. We apply transciphering technology and an efficient Fully Homomorphic Encryption (FHE) cryptosystem to optimize overall performance. Simultaneously, a Computing Server and a Trusted Third Party (TTP) are adopted to perform all the homomorphic operations on the server side, leaving clients to encrypt and decrypt local data using stream ciphers. Then, we implement an instance of our scheme based on FiLIP stream cipher, a homomorphic comparison algorithm, and a homomorphic traverse algorithm. The experimental results show that the various overheads on the client side are relatively lower.

关键词： Performance evaluation Ciphers Privacy Machine learning algorithms Federated learning Scalability Servers

来源：评论

学校读者我要写书评

暂无评论

ATA-Cache: Contention Mitigation for GPU Shared L1 Cache with Aggregated Tag Array

arXiv

引用

arXiv 2023年

作者： Xu, Xiangrong Wang, Liang Xiao, Limin Liu, Lei Xie, Xilong Han, Meng Liu, Hao State Key Laboratory of Software Development Environment School of Computer Science and Engineering Beihang University Beijing100191 China

GPU shared L1 cache is a promising architecture while still suffering from high resource contentions. We present a GPU shared L1 cache architecture with an aggregated tag array that minimizes the L1 cache contentions and takes full advantage of inter-core locality. The key idea is to decouple and aggregate the tag arrays of multiple L1 caches so that the cache requests can be compared with all tag arrays in parallel to probe the replicated data in other caches. The GPU caches are only accessed by other GPU cores when replicated data exists, filtering out unnecessary cache accesses that cause high resource contentions. The experimental results show that GPU IPC can be improved by 12% on average for applications with a high inter-core locality. Copyright © 2023, The Authors. All rights reserved.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Federated Learning based Path Planning Method for Crowd Evacuation 25

Federated Learning based Path Planning Method for Crowd Evac...

引用

25th IEEE International Conference on computer Supported Cooperative Work in Design, CSCWD 2022

作者： Yuan, Zihao Lu, Dianjie Zhang, Guijuan Liu, Hong School of Information Science and Engineering Shandong Normal University Shandong Provincial Key Laboratory for Novel Distributed Computer Software Technology State Key Laboratory of High-end Server and Storage Technology Shandong Jinan China

ISBN: (纸本)9781665405270

The surveillance videos captured by cameras deployed in large-scale places can provide help for path planning of crowd evacuation in emergencies. However, most surveillance videos are not allowed to be shared in order to avoid privacy disclosure. Thus, it is difficult to obtain the video data of the whole scenario for path planning which greatly reduces effectiveness of path planning. How to provide global path planning for crowd evacuation while ensuring privacy protection is a challenging problem. To solve this problem, we propose a federated learning based path planning method for crowd evacuation. In this method, the potential field model are introduced into the federated learning framework to provide the privacy protection since only the potential field information is aggregated between the camera terminal and the central server, instead of the surveillance videos. First, we construct the local potential field for the surveillance scene of each camera by extracting the scene information and the crowds information. Second, the central server periodically performs the global potential field information aggregation and integrates these information to plan the optimal global path dynamically for crowd evacuation. Finally, we implemente a simulation platform to simulate the process of crowd movement and visualize the dynamic change of the potential energy field. Experimental results prove that the proposed method can provide privacy-preserving path planning for multi-camera evacuation scenarios. © 2022 IEEE.

关键词： Motion planning

来源：评论

学校读者我要写书评

暂无评论

Blockchain-based Crowd-sensing Trust Management Mechanism for Crowd Evacuation 25

Blockchain-based Crowd-sensing Trust Management Mechanism fo...

引用

25th IEEE International Conference on computer Supported Cooperative Work in Design, CSCWD 2022

作者： Yu, Jiawen Zhang, Guijuan Lu, Dianjie Liu, Hong School of Information Science and Engineering Shandong Normal University Shandong Provincial Key Laboratory for Novel Distributed Computer Software Technology State Key Laboratory of High-end Server and Storage Technology Shandong Jinan China

ISBN: (纸本)9781665405270

With the continuous urban expansion, the safety of public places has attracted more and more attention. Once an accident occurs and people distrust their surroundings, it will cause congestion, trampling and other events. Individual trust value is particularly important in crowd evacuation. However, it is very difficult to evaluate the individual credibility with existing methods. We propose a blockchain-based crowd-sensing trust management mechanism (B-CSTM) to address this problem. First, we provide a message credibility evaluation mechanism that can evaluate the credibility of messages sent by individuals on spot. Second, we developed a trust calculation method that uses a distributed Evacuation Perception Unit (EPU) to calculate individuals' trust values by perception scores. Finally, we build a blockchain-based trust management mechanism model that uses blockchain to store and query trust values. Through experimental analysis, we visualize the trust results in the evacuation scenario. It is shown that the mechanism is feasible for the management of trust values (collection, calculation and storage, etc.) in crowd evacuation. © 2022 IEEE.

关键词： Blockchain

来源：评论

学校读者我要写书评

暂无评论

Improving Local Search for Pseudo Boolean Optimization by Fragile Scoring Function and Deep Optimization 29

Improving Local Search for Pseudo Boolean Optimization by Fr...

引用

29th International Conference on Principles and Practice of Constraint Programming, CP 2023

作者： Zhou, Wenbo Zhao, Yujiao Wang, Yiyuan Cai, Shaowei Wang, Shimao Wang, Xinyu Yin, Minghao School of Information Science and Technology Northeast Normal University Changchun China Key Laboratory of Applied Statistics of MOE Northeast Normal University Changchun China Key Laboratory of Symbolic Computation and Knowledge Engineering of MOE Jilin University Changchun China State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences Beijing China School of Computer Science and Technology University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9783959773003

Pseudo-Boolean optimization (PBO) is usually used to model combinatorial optimization problems, especially for some real-world applications. Despite its significant importance in both theory and applications, there are few works on using local search to solve PBO. This paper develops a novel local search framework for PBO, which has three main ideas. First, we design a two-level selection strategy to evaluate all candidate variables. Second, we propose a novel deep optimization strategy to disturb some search spaces. Third, a sampling flipping method is applied to help the algorithm jump out of local optimum. Experimental results show that the proposed algorithms outperform three state-of-the-art PBO algorithms on most instances. © Wenbo Zhou, Yujiao Zhao, Yiyuan Wang, Shaowei Cai, Shimao Wang, Xinyu Wang, and Minghao Yin.

关键词： Combinatorial optimization

来源：评论

学校读者我要写书评

暂无评论

Data-Driven Crowd Motion Modeling With Group Properties Analysis 25

Data-Driven Crowd Motion Modeling With Group Properties Anal...

引用

25th IEEE International Conference on computer Supported Cooperative Work in Design, CSCWD 2022

作者： Jia, Chuanhua Lu, Dianjie Zhang, Guijuan Liu, Hong School of Information Science and Engineering Shandong Normal University Shandong Provincial Key Laboratory for Novel Distributed Computer Software Technology State Key Laboratory of High-end Server and Storage Technology Shandong Jinan China

ISBN: (纸本)9781665405270

Traditional crowd evacuation simulation methods focus on the analysis of individual behavior in the crowd, but ignore the analysis of potential group properties in the process of crowd motion, which reduces the visual realism of the crowd evacuation simulation process. During crowd motion, crowds cause unconscious self-organization due to common destinations and social relationships, thus forming groups. Considering the influence of group properties on crowd motion is important to improve the visual realism of crowd motion simulation. To address this problem, we propose a crowd motion modeling method for group properties analysis (GPA-CMM). First, we build a data-driven group properties quantification (DGPQ) model to describe the characteristics of group motion accurately. In the model, we divide the crowd into several groups, extract motion properties of the crowd, and quantify the features of intra-group stability and inter-group conflict. Then, in order to analyze their influences of group properties, we build a stability and conflict based crowd motion analysis (SC-CMA) model. Finally, we implement a crowd simulation system based on SC-CMA, visualize the results of the theoretical analysis in a graphical way. The experimental results show that the method can simulate crowd motion more realistically. © 2022 IEEE.

关键词： Behavioral research

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Teaching and Learning of Representations on Graphs 23

Self-Supervised Teaching and Learning of Representations on ...

引用

2023 World Wide Web Conference, WWW 2023

作者： Wan, Liangtian Fu, Zhenqiang Sun, Lu Wang, Xianpeng Xu, Gang Yan, Xiaoran Xia, Feng Key Laboratory for Ubiquitous Network Service Software of Liaoning Province School of Software Dalian University of Technology Dalian China Department of Communication Engineering Institute of Information Science Technology Dalian Maritime University Dalian China State Key Laboratory of Marine Resource Utilization in South China Sea School of Information and Communication Engineering Hainan University Haikou China State Key Laboratory of Millimeter Waves School of Information Science and Engineering Southeast University Nanjing China Research Center of Big Data Intelligence Research Institute of Artificial Intelligence Zhejiang Lab Hangzhou China School of Computing Technologies Rmit University Melbourne Australia

ISBN: (纸本)9781450394161

Recent years have witnessed significant advances in graph contrastive learning (GCL), while most GCL models use graph neural networks as encoders based on supervised learning. In this work, we propose a novel graph learning model called GraphTL, which explores self-supervised teaching and learning of representations on graphs. One critical objective of GCL is to retain original graph information. For this purpose, we design an encoder based on the idea of unsupervised dimensionality reduction of locally linear embedding (LLE). Specifically, we map one iteration of the LLE to one layer of the network. To guide the encoder to better retain the original graph information, we propose an unbalanced contrastive model consisting of two views, which are the learning view and the teaching view, respectively. Furthermore, we consider the nodes that are identical in muti-views as positive node pairs, and design the node similarity scorer so that the model can select positive samples of a target node. Extensive experiments have been conducted over multiple datasets to evaluate the performance of GraphTL in comparison with baseline models. Results demonstrate that GraphTL can reduce distances between similar nodes while preserving network topological and feature information, yielding better performance in node classification. © 2023 ACM.

关键词： Classification (of information)

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：