检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Zhang, Hai Yu, Hang Zhao, Junqiao Zhang, Di Huang, Chang Zhou, Hongtu Zhang, Xiao Ye, Chen Department of Computer Science Tongji University Shanghai China MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China

Designing and deriving effective model-based reinforcement learning (MBRL) algorithms with a performance improvement guarantee is challenging, mainly attributed to the high coupling between model learning and policy optimization. Many prior methods that rely on return discrepancy to guide model learning ignore the impacts of model shift, which can lead to performance deterioration due to excessive model updates. Other methods use performance difference bound to explicitly consider model shift. However, these methods rely on a fixed threshold to constrain model shift, resulting in a heavy dependence on the threshold and a lack of adaptability during the training process. In this paper, we theoretically derive an optimization objective that can unify model shift and model bias and then formulate a fine-tuning process. This process adaptively adjusts the model updates to get a performance improvement guarantee while avoiding model overfitting. Based on these, we develop a straightforward algorithm USB-PO2 (Unified model Shift and model Bias Policy Optimization). Empirical results show that USB-PO achieves state-of-the-art performance on several challenging benchmark tasks. © 2023, CC BY.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

Focus On What Matters: Separated Models For Visual-Based RL Generalization

arXiv

引用

arXiv 2024年

作者： Zhang, Di Lv, Bowen Zhang, Hai Yang, Feifan Zhao, Junqiao Yu, Hang Huang, Chang Zhou, Hongtu Ye, Chen Jiang, Changjun Department of Computer Science Tongji University Shanghai China MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China

A primary challenge for visual-based Reinforcement Learning (RL) is to generalize effectively across unseen environments. Although previous studies have explored different auxiliary tasks to enhance generalization, few adopt image reconstruction due to concerns about exacerbating overfitting to task-irrelevant features during training. Perceiving the pre-eminence of image reconstruction in representation learning, we propose SMG (Separated Models for Generalization), a novel approach that exploits image reconstruction for generalization. SMG introduces two model branches to extract task-relevant and task-irrelevant representations separately from visual observations via cooperatively reconstruction. Built upon this architecture, we further emphasize the importance of task-relevant features for generalization. Specifically, SMG incorporates two additional consistency losses to guide the agent’s focus toward task-relevant areas across different scenarios, thereby achieving free from overfitting. Extensive experiments in DMC demonstrate the SOTA performance of SMG in generalization, particularly excelling in video-background settings. Evaluations on robotic manipulation tasks further confirm the robustness of SMG in real-world applications. Source code is available at https://***/r/SMG/. © 2024, CC BY.

关键词： 3D reconstruction

来源：评论

学校读者我要写书评

暂无评论

Automatic Web service composition based on Horn clauses and Petri nets

引用

Expert systems with Applications 2011年第10期38卷 13024-13031页

作者： Tang, Xianfei Jiang, Changjun Zhou, Mengchu Department of Computer Science and Technology Tongji University Shanghai 201804 China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai 201804 China Department of Electrical and Computer Engineering New Jersey Institute of Technology Newark NJ 07102 United States

This paper introduces an automatic Web service composition method based on logical inference of Horn clauses and Petri nets. The Web service composition problem is transformed into the logical inference problem of Horn clauses by exploring the dependency relations among services. An algorithm is presented, to determine whether the composite service exists in time that is linear in the size of the knowledge base. It can also return the rule set necessary for the composition. Petri nets are chosen to model the rule set and its structural analysis techniques are used to obtain the composite service. © 2010 Elsevier Ltd. All rights reserved.

关键词： Automatic reasoning Knowledge base Petri nets service composition Web service

来源：评论

学校读者我要写书评

暂无评论

Process-aware service composition based on Petri net

引用

Journal of Harbin Institute of Technology (New Series) 2008年第SUPPL.期15卷 162-166页

作者： Sun, Ping Li, Xiangmei Department of Computer Science and Engineering Tongji University Shanghai 201804 China Embedded System and Service Computing Key Laboratory Shanghai 201804 China

One of the most promising advantages of Web service technology is the possibility of creating value-added services by combining existing ones. A major challenge is how to discover and select concrete service according to user requirements. This paper addresses the topic of service discovery composite Web services. The main feature is that we take the process model as well as service profile into account. Firstly, the process models of Web services are translated into Petri nets. Based on this, we propose a service matchmaking algorithm, via comparing the functionality compatibility and process consistency, thus leading to more accurate matchmaking.

关键词： Web services

来源：评论

学校读者我要写书评

暂无评论

A robust facial descriptor for face recognition

A robust facial descriptor for face recognition

引用

International Conference on Applications and Techniques in Cyber Security and Intelligence, ATCSI 2017

作者： Liu, Na Duan, Huixian Song, Lei Yan, Zhiguo Cyber Physical System R&D Center The Third Research Institute of Ministry of Public Security Shanghai China The Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai China Shanghai International Technology and Trade United Co. Ltd Shanghai China

ISBN: (纸本)9783319670706

Illumination, occlusion, pose and expression variations are the most common challenging problems for face recognition in many real-world applications. However, existing face recognition methods are proposed to handle part of these variations. In this paper, we propose a robust facial descriptor to address this issue. First, we apply a chain of three processing to tackle the illumination variation. Second, we compute the facial sparse local descriptor to handle the occlusion, pose, and expression variations. Experimental evaluation on the FRGC database shows that our approach is able to achieve very promising recognition rates under uncontrolled environments. © 2018, Springer International Publishing AG.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

How to fine-tune the model: unified model shift and model bias policy optimization 23

How to fine-tune the model: unified model shift and model bi...

引用

Proceedings of the 37th International Conference on Neural Information Processing systems

作者： Hai Zhang Hang Yu Junqiao Zhao Di Zhang Chang Huang Hongtu Zhou Xiao Zhang Chen Ye Department of Computer Science Tongji University Shanghai China and MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China

Designing and deriving effective model-based reinforcement learning (MBRL) algorithms with a performance improvement guarantee is challenging, mainly attributed to the high coupling between model learning and policy optimization. Many prior methods that rely on return discrepancy to guide model learning ignore the impacts of model shift, which can lead to performance deterioration due to excessive model updates. Other methods use performance difference bound to explicitly consider model shift. However, these methods rely on a fixed threshold to constrain model shift, resulting in a heavy dependence on the threshold and a lack of adaptability during the training process. In this paper, we theoretically derive an optimization objective that can unify model shift and model bias and then formulate a fine-tuning process. This process adaptively adjusts the model updates to get a performance improvement guarantee while avoiding model over-fitting. Based on these, we develop a straightforward algorithm USB-PO (Unified model Shift and model Bias Policy Optimization). Empirical results show that USB-PO achieves state-of-the-art performance on several challenging benchmark tasks. Code: https://***/betray12138/***

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Cloudedge-Combined Data Placement Strategy Based on User Access Regions 6

A Cloudedge-Combined Data Placement Strategy Based on User A...

引用

6th International Conference on Big Data and Information Analytics, BigDIA 2020

作者： Cao, Eryang Wang, Pengwei Yan, Chungang Jiang, Changjun Department of Computer Science and Technology Tongji University Shanghai200092 China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai200092 China School of Computer Science and Technology Donghua University Shanghai201620 China

To improve the availability of data in the cloud and avoid vendor lock-in risk, multi-cloud storage is attracting more and more attentions. However, accessing data from the cloud usually has some disadvantages such as higher latency and more bandwidth-consuming. Edge computing, which has become popular in recent years, could effectively solve the above problems while it usually reduces availability. Storage only in the cloud or edge environment is limited by many factors. To this end, we build mathematical model which is abstracted from real access mode of users in consideration of multi-cloud and edge. This paper presents a method combined NSGAII with multi-group method which has better ability of global search to help users determine cloud and edge services to store and access data object. A large number of experiments using real service provider information demonstrate that the proposed algorithm outperforms existing methods and effectiveness of it. © 2020 IEEE.

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Biped robot walking using central pattern generator and genetic algorithm

Biped robot walking using central pattern generator and gene...

引用

IASTED International Conference on Robotics, Robo 2010

作者： Liu, Chengju Chen, Qijun Wang, Danwei School of Electronics and Information Engineering Tongji University Shanghai 201804 China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai 201804 China School of Electrical and Electronic Engineering Nanyang Technological University Singapore Singapore

ISBN: (纸本)9780889868595

Central pattern generator (CPG) plays an important role in rhythmic activities of animals and this mechanism is an important inspiration source for the motion control of legged robots. In this paper, by using CPGs and function mapping mechanism, a high-efficiency distributed CPG control network is constructed to realize the locomotion control of biped NAO robot. To realize stable and coordinated locomotion, the parameters of the CPG network are evolved by multi-object genetic algorithm (MOGA). Simulations with Webots validate the feasibility and efficiency of the presented CPG-based control method.

关键词： Genetic algorithms

来源：评论

学校读者我要写书评

暂无评论

A speed estimation method of vehicles based on road monitoring video-images

A speed estimation method of vehicles based on road monitori...

引用

International Conference on Applications and Techniques in Cyber Security and Intelligence, ATCSI 2017

作者： Huixian, Duan Jun, Wang Lei, Song Yixin, Zhao Liu, Na The Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai China Cyber Physical System R&D Center The Third Research Institute of Ministry of Public Security Shanghai China Shanghai International Technology and Trade United Co. Ltd Shanghai China

ISBN: (纸本)9783319670706

In order to reduce traffic accidents and road congestion in many cities, vehicle speed estimation is very critical and important to observe speed limitation law and traffic conditions. In this paper, we present a speed estimation method of vehicles based on road monitoring video-images. Firstly, we set up a word coordinate system on the license plate in one vehicle image. Next, for small vehicles in China, according to the known length and width of the license plate, the spatial transformation matrix between the word coordinate system and the image coordinate system is derived. Then, based on the spatial transform matrix, compute the corner spatial coordinates of the license plate in each frame, and then estimate the vehicle speed. Finally, experimental results on read data have shown that the vehicle speed can be estimated within the acceptable error range (±3 km/h), and then have demonstrated the effectiveness of the proposed vehicle speed estimation method. © 2018, Springer International Publishing AG.

关键词： Speed

来源：评论

学校读者我要写书评

暂无评论

Composite behavioral modeling for identity theft detection in online social networks

arXiv

引用

arXiv 2018年

作者： Wang, Cheng Yang, Bo Department of Computer Science and Technology Tongji University Key Laboratory of Embedded System and Service Computing Ministry of Education China

In this work, we aim at building a bridge from poor behavioral data to an effective, quick-response, and robust behavior model for online identity theft detection. We concentrate on this issue in online social networks (OSNs) where users usually have composite behavioral records, consisting of multi-dimensional low-quality data, e.g., offline check-ins and online user generated content (UGC). As an insightful result, we find that there is a complementary effect among different dimensions of records for modeling users' behavioral patterns. To deeply exploit such a complementary effect, we propose a joint model to capture both online and offline features of a user's composite behavior. We evaluate the proposed joint model by comparing with some typical models on two real-world datasets: Foursquare and Yelp. In the widely-used setting of theft simulation (simulating thefts via behavioral replacement), the experimental results show that our model outperforms the existing ones, with the AUC values 0:956 in Foursquare and 0:947 in Yelp, respectively. Particularly, the recall (True Positive Rate) can reach up to 65:3% in Foursquare and 72:2% in Yelp with the corresponding disturbance rate (False Positive Rate) below 1%. It is worth mentioning that these performances can be achieved by examining only one composite behavior (visiting a place and posting a tip online simultaneously) per authentication, which guarantees the low response latency of our method. This study would give the cybersecurity community new insights into whether and how a real-time online identity authentication can be improved via modeling users' composite behavioral patterns. Copyright © 2018, The Authors. All rights reserved.

关键词： Behavioral research

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：