检索结果-内蒙古大学图书馆

Provably Convergent Federated Trilevel Learning

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Jiao, Yang Yang, Kai Wu, Tiancheng Jian, Chengtao Huang, Jianwei Department of Computer Science and Technology Tongji University China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University China Shanghai Research Institute for Intelligent Autonomous Systems China School of Science and Engineering The Chinese University of Hong Kong Shenzhen China Shenzhen Institute of Artificial Intelligence and Robotics for Society China

Trilevel learning, also called trilevel optimization (TLO), has been recognized as a powerful modelling tool for hierarchical decision process and widely applied in many machine learning applications, such as robust neural architecture search, hyperparameter optimization, and domain adaptation. Tackling TLO problems has presented a great challenge due to their nested decision-making structure. In addition, existing works on TLO face the following key challenges: 1) they all focus on the non-distributed setting, which may lead to privacy breach;2) they do not offer any non-asymptotic convergence analysis which characterizes how fast an algorithm converges. To address the aforementioned challenges, this paper proposes an asynchronous federated trilevel optimization method to solve TLO problems. The proposed method utilizes µ-cuts to construct a hyper-polyhedral approximation for the TLO problem and solve it in an asynchronous manner. We demonstrate that the proposed µ-cuts are applicable to not only convex functions but also a wide range of non-convex functions that meet the µ-weakly convex assumption. Furthermore, we theoretically analyze the non-asymptotic convergence rate for the proposed method by showing its iteration complexity to obtain ϵ-stationary point is upper bounded by (Formula Presented). Extensive experiments on real-world datasets have been conducted to elucidate the superiority of the proposed method, e.g., it has a faster convergence rate with a maximum acceleration of approximately 80%. Copyright © 2023, The Authors. All rights reserved.

关键词： Decision making

Solving Colored Traveling Salesman Problem via Multi-neighborhood Simulated Annealing Search 18

学校读者我要写书评

暂无评论

Solving Colored Traveling Salesman Problem via Multi-neighbo...

18th IEEE International Conference on Networking, Sensing and Control, ICNSC 2021

作者： Zhou, Yangming Xu, Wenqiang Fu, Zhang-Hua Zhou, Mengchu East China University of Science and Technology Department of Computer Science and Engineering Shanghai200237 China Macau Institute of Systems Engineering Macau University of Science and Technology 999078 China Shenzhen Institute of Artificial Intelligence and Robotics for Society Shenzhen518172 China Institute of Robotics and Intelligent Manufacturing The Chinese University of Hong Kong Shenzhen Shenzhen518172 China New Jersey Institute of Technology Department of Electrical and Computer Engineering NewarkNJ07102 United States Department of Cyber-Physical Systems St. Petersburg State Marine Technical University St. Petersburg198262 Russia

ISBN: (纸本)9781665440486

A colored traveling salesman problem (CTSP) is an important variant of the well-known multiple traveling salesman problem, which uses colors to differentiate salesmen's accessibility to individual cities to be visited. As a highly useful model for some complex scheduling problems, CTSP is NP-hard. A Multi-neighborhood Simulated Annealing Search (MSAS) approach is proposed to solve it in this paper. Starting from an initial solution, it iterates through two complementary neighborhoods: intra-route and inter-route neighborhoods. Experiments on three groups of 60 widely-used benchmark instances show that it achieves highly competitive performance compared to state-of-the-art algorithms. Moreover, MSAS can be integrated into other search methods to further improve performance, which is demonstrated by using a recently proposed iterated two-phase local search. © 2021 IEEE.

关键词： Traveling salesman problem

Deep Learning-Based Prediction of Subsurface Oil Reservoir Pressure Using Spatio-Temporal Data

学校读者我要写书评

暂无评论

Deep Learning-Based Prediction of Subsurface Oil Reservoir P...

Annual Conference of Industrial Electronics Society

作者： Haibo Cheng Yunpeng He Peng Zeng Shichao Li Valeriy Vyatkin State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China Key Laboratory of Networked Control Systems Chinese Academy of Sciences Shenyang China Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang China University of Chinese Academy of Sciences Beijing China Department of Computer Science Electrical and Space Engineering Luleå University of Technology Luleå Sweden Department of Electrical Engineering and Automation Aalto University Helsinki Finland

Prediction of subsurface oil reservoir pressure are critical to hydrocarbon production. However, the accurate pressure estimation faces great challenges due to the complexity and uncertainty of reservoir. The underground seepage flow and petrophysical parameters (permeability and porosity) are important but difficult to measure in oilfield. Deep learning methods have been successfully used in reservoir engineering and oil & gas production process. In this study, the effective but inaccessible subsurface seepage fields are not used, only the spatial coordinates and temporal information are selected as model input to predict reservoir pressure. A stacked GRU-based deep learning model is proposed to map the relationship between spatio-temporal data and reservoir pressure. The proposed deep learning method is verified by using a three-dimensional reservoir model, and compared with commonly-used methods. The results show that the stacked GRU model has a better performance and higher accuracy than other deep learning or machine learning methods in pressure prediction.

关键词：

AI-fuzzy markup language with computational intelligence for high-school student learning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Lee, Chang-Shing Wang, Mei-Hui Nojima, Yusuke Reformat, Marek Guo, Leo Computer Science and Information Engineering Department National University of Tainan Taiwan Computer Science and Intelligent Systems Department Osaka Prefecture University Japan Electrical and Computer Engineering Department University of Alberta Canada Nuwa Robotics Taiwan

Computational Intelligence (CI), which includes fuzzy logic (FL), neural network (NN), and evolutionary computation (EC), is an imperative branch of artificial intelligence (AI). As a core technology of AI, it plays a vital role in developing intelligent systems, such as games and game engines, neural-based systems including a variety of deep network models, evolutionarybased optimization methods, and advanced cognitive techniques. The 2021 IEEE CIS Summer School on CI for High-School Student Learning was held physically at the JanFuSun Resort Hotel, Taiwan, and virtually on Zoom, on August 10-12, 2021. The main contents of the Summer School were lectures focused on the basics of FL, NN, and EC and the workshop on AIoT (Artificial Intelligence of Things). Invited speakers gave nine courses covering topics like CI real-world applications, fundamentals of FL, and the introduction to NN and EC. The 2021 Summer School was supported by the 2021 IEEE CIS High School Outreach Subcommittee. We also invited students and teachers of high and elementary schools from Taiwan, Japan, and Indonesia. They attended the school and participated in AIoT workshop, gaining experience in applications of AIoT-FML learning tools. According to the short report and feedback from the involved students and teachers, we find out that most participants have quickly understood the principles of CI, FL, NN, and EC. In addition, one of the teachers sent the following remark to the organizers: "This is a great event to introduce students to computational intelligence at a young age, stimulate them to be involved in rapidly evolving fields, and foster participation in future research adventures.". Copyright © 2021, The Authors. All rights reserved.

关键词： intelligent systems

CARL: Camera-Agnostic Representation Learning for Spectral Image Analysis

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Baumann, Alexander Ayala, Leonardo Seidlitz, Silvia Sellner, Jan Studier-Fischer, Alexander Özdemir, Berkin Maier-Hein, Lena Ilic, Slobodan Siemens AG Munich Germany Heidelberg Germany Medical Faculty Faculty of Mathematics and Computer Science Heidelberg University Germany NCT Heidelberg Germany HIDSS4Health Heidelberg Germany Department of Urology and Urosurgery University Medical Center Mannheim Germany Department of General Visceral and Transplantation Surgery Heidelberg University Hospital Germany Division of Intelligent Systems and Robotics in Urology DKFZ Heidelberg Germany DKFZ Hector Cancer Institute University Medical Center Mannheim Germany

Spectral imaging offers promising applications across diverse domains, including medicine and urban scene understanding, and is already established as a critical modality in remote sensing. However, variability in channel dimensionality and captured wavelengths among spectral cameras impede the development of AI-driven methodologies, leading to camera-specific models with limited generalizability and inadequate cross-camera applicability. To address this bottleneck, we introduce CARL, a model for Camera-Agnostic Representation Learning across RGB, multispectral, and hyperspectral imaging modalities. To enable the conversion of a spectral image with any channel dimensionality to a camera-agnostic embedding, we introduce wavelength positional encoding and a self-attention-cross-attention mechanism to compress spectral information into learned query representations. Spectral-spatial pre-training is achieved with a novel spectral self-supervised JEPA-inspired strategy tailored to CARL. Large-scale experiments across the domains of medical imaging, autonomous driving, and satellite imaging demonstrate our model’s unique robustness to spectral heterogeneity, outperforming on datasets with simulated and real-world cross-camera spectral variations. The scalability and versatility of the proposed approach position our model as a backbone for future spectral foundation models. © 2025, CC BY.

关键词： Image coding

F2A2: flexible fully-decentralized approximate actor-critic for cooperative multi-agent reinforcement learning

学校读者我要写书评

暂无评论

The Journal of Machine Learning Research 2023年第1期24卷 8447-8521页

作者： Wenhao Li Bo Jin Xiangfeng Wang Junchi Yan Hongyuan Zha School of Data Science The Chinese University of Hong Kong Shenzhen Shenzhen Institute of Artificial Intelligence and Robotics for Society Shenzhen China School of Software Engineering Shanghai Research Institute for Intelligent Autonomous Systems Tongji University Shanghai China School of Computer Science and Technology Key Laboratory of Mathematics and Engineering Applications Ministry of Education East China Normal University Shanghai China Department of Computer Science and Engineering Key Laboratory of Artificial Intelligence Ministry of Education Shanghai Jiao Tong University Shanghai China

Traditional centralized multi-agent reinforcement learning (MARL) algorithms are sometimes unpractical in complicated applications due to non-interactivity between agents, the curse of dimensionality, and computation complexity. Hence, several decentralized MARL algorithms are motivated. However, existing decentralized methods only handle the fully cooperative setting where massive information needs to be transmitted in training. The block coordinate gradient descent scheme they used for successive independent actor and critic steps can simplify the calculation, but it causes serious bias. This paper proposes a exible fully decentralized actor-critic MARL framework, which can combine most of the actor-critic methods and handle large-scale general cooperative multi-agent settings. A primal-dual hybrid gradient descent type algorithm framework is designed to learn individual agents separately for decentralization. From the perspective of each agent, policy improvement and value evaluation are jointly optimized, which can stabilize multi-agent policy learning. Furthermore, the proposed framework can achieve scalability and stability for the large-scale environment. This framework also reduces information transmission by the parameter sharing mechanism and novel modeling-other-agents methods based on theory-of-mind and online supervised learning. Sufficient experiments in cooperative Multi-agent Particle Environment and StarCraft II show that the proposed decentralized MARL instantiation algorithms perform competitively against conventional centralized and decentralized methods.

关键词： cooperative MARL decentralized actor-critic primal-dual method

Long-Horizon Multi-Robot Rearrangement Planning for Construction Assembly

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Hartmann, Valentin N. Orthey, Andreas Driess, Danny Oguz, Ozgur S. Toussaint, Marc Machine Learning & Robotics Lab University of Stuttgart Germany Learning and Intelligent Systems Group TU Berlin Germany Department of Computer Engineering Bilkent University Turkey

Robotic assembly planning enables architects to explicitly account for the assembly process during the design phase, and enables efficient building methods that profit from the robots' different capabilities. Previous work has addressed planning of robot assembly sequences and identifying the feasibility of architectural designs. This paper extends previous work by enabling planning with large, heterogeneous teams of robots. We present a planning system which enables parallelization of complex task and motion planning problems by iteratively solving smaller subproblems. Combining optimization methods to solve for manipulation constraints with a sampling-based bi-directional space-time path planner enables us to plan cooperative multirobot manipulation with unknown arrival-times. Thus, our solver allows for completing subproblems and tasks with differing timescales and synchronizes them effectively. We demonstrate the approach on multiple case-studies to show the robustness over long planning horizons and scalability to many objects and agents of our algorithm. Finally, we also demonstrate the execution of the computed plans on two robot arms to showcase the feasibility in the real world. Copyright © 2021, The Authors. All rights reserved.

关键词： robotics

Theory of Mind Abilities Predict Robot's Gaze Effects on Object Preference

学校读者我要写书评

暂无评论

IEEE Transactions on Affective Computing 2025年

作者： Manzi, Federico Ishikawa, Mitsuhiko Dio, Cinzia Di Itakura, Shoji Kanda, Takayuki Ishiguro, Hiroshi Massaro, Davide Marchetti, Antonella Università Cattolica del Sacro Cuore Research Unit on Theory of Mind Department of Psychology Milan Italy Università Cattolica del Sacro Cuore Research Unit in Psychology and Robotics in the Lifespan Department of Psychology Milan Italy IRCCS Fondazione Don Carlo Gnocchi Milan Italy Hitotsubashi University Hitotsubashi Institute for Advanced Study Japan Ritsumeikan University Kyoto Japan Kyoto University Department of Computer Science Human-Robot Interaction Laboratory Kyoto Japan IRC'HIL Keihana Science City Advanced Telecommunications Research Institute International Kyoto Japan Osaka University Department of Intelligent Systems Toyonaka Japan

This study investigated the differences between human and robot gaze in influencing preference formation, and examined the role of Theory of Mind (ToM) abilities in this process. Human eye gaze is one of the most important sources of information for social interaction and research has demonstrated its effectiveness in influencing people's preference. With increasing technological development, we will interact with robots that can exhibit gaze behavior and influence people's preference. It is unclear whether there are any differences between humans and robots in this process. The present study aimed to analyze the role of the gaze of a robot and a human in influencing the ascription of a preference to the gazer and the participants' preference. Furthermore, we examined the role of ToM abilities in preference formation. The results showed that the gaze has a greater effect on the gazer preference compared to participants' preference regardless of the agent (human or robot). In addition, ToM abilities predict both gazer and individual preferences in the robot's condition only even though different socio-cognitive mechanisms are involved. The study suggests that adults are cognitively able to process the gaze of a robot similar to a human, recognizing the underlying mental state. However, only for the robot, different cognitive mechanisms are involved in the gazer (i.e., perspective taking) and participants' preference formation (i.e., advanced ToM). © 2010-2012 IEEE.

关键词： Social robots

Automated Food Intake Monitoring System to Prevent Malnutrition Using the Tiago Robot Camera*

学校读者我要写书评

暂无评论

Automated Food Intake Monitoring System to Prevent Malnutrit...

Data Science and Engineering in Healthcare, Medicine and Biology, IEEE EMBS Special Topic Conference on

作者： Fotios S. Konstantakopoulos Daphni K. Plati Labrina A. Pliakou Francesco Scotto Di Luzio Nevio Luigi Tagliamonte Loredana Zollo Manolis Tsiknakis Dimitrios I. Fotiadis Unit of Medical Technology and Intelligent Information Systems Materials Science and Engineering Department University of Ioannina Biomedical Research Institute FORTH Ioannina GR 45110 Greece Research Unit of Advanced Robotics and Human-Centred Technologies Campus Bio-Medico University Rome Italy Institute of Computer Science FORTH Heraklion Crete GR 70013 Greece

The management of a daily diet is a significant concern among individuals in modern culture. The utilization of dietary assessment systems has significantly contributed to the efficient management of malnutrition and dietary habits over a period of time. In order to determine the nutritional value of the food that individuals consume, this study introduces a novel food monitoring system and its architecture. By creating a dataset of food images in a tray and using image processing, machine learning, and computer vision techniques, the nutrients and calories consumed by a patient are calculated. The proposed system achieves a 95.6% Intersection over Union score in the segmentation task, 92.3% top-1 accuracy in the classification task, and 4.6 g mean weight absolute error in the weight estimation *** Relevance: The proposed system has the ability to calculate the nutritional composition, encompassing calories, proteins, fats, and carbs, of food consumed by a patient, making it suitable for nutritional monitoring applications and healthcare systems that aim to monitor malnutrition in hospital patients.

关键词：