检索结果-内蒙古大学图书馆

Multimodal Large Models Are Effective Action Anticipators

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Wang, Binglu Tian, Yao Wang, Shunzhou Yang, Le College of Information and Control Engineering Xi'an University of Architecture and Technology Xi'An710055 China The School of Astronautics Northwestern Polytechnical University Xi'An710072 China School of Electronic and Computer Engineering Shenzhen Graduate School Peking University Shenzhen518055 China Institute of Artificial Intelligence Hefei Comprehensive National Science Center China The School of Information Science and Technology University of Science and Technology of China Hefei230022 China

The task of long-term action anticipation demands solutions that can effectively model temporal dynamics over extended periods while deeply understanding the inherent semantics of actions. Traditional approaches, which primarily rely on recurrent units or Transformer layers to capture long-term dependencies, often fall short in addressing these challenges. Large Language Models (LLMs), with their robust sequential modeling capabilities and extensive commonsense knowledge, present new opportunities for long-term action anticipation. In this work, we introduce the ActionLLM framework, a novel approach that treats video sequences as successive tokens, leveraging LLMs to anticipate future actions. Our baseline model simplifies the LLM architecture by setting future tokens, incorporating an action tuning module, and reducing the textual decoder layer to a linear layer, enabling straightforward action prediction without the need for complex instructions or redundant descriptions. To further harness the commonsense reasoning of LLMs, we predict action categories for observed frames and use sequential textual clues to guide semantic understanding. In addition, we introduce a Cross-Modality Interaction Block, designed to explore the specificity within each modality and capture interactions between vision and textual modalities, thereby enhancing multimodal tuning. Extensive experiments on benchmark datasets demonstrate the superiority of the proposed ActionLLM framework, encouraging a promising direction to explore LLMs in the context of action anticipation. Code is available at https://***/2tianyao1/***. Copyright © 2025, The Authors. All rights reserved.

关键词： Semantics

Do We Really Need to Use Constraint Violation in Constrained Evolutionary Multi-Objective Optimization?

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Shuang Li, Ke Li, Wei Control and Simulation Center Harbin Institute of Technology China Department of Computer Science University of Exeter ExeterEX4 4QF United Kingdom

Constraint violation has been a building block to design evolutionary multi-objective optimization algorithms for solving constrained multi-objective optimization problems. However, it is not uncommon that the constraint violation is hardly approachable in real-world black-box optimization scenarios. It is unclear that whether the existing constrained evolutionary multi-objective optimization algorithms, whose environmental selection mechanism are built upon the constraint violation, can still work or not when the formulations of the constraint functions are unknown. Bearing this consideration in mind, this paper picks up four widely used constrained evolutionary multi-objective optimization algorithms as the baseline and develop the corresponding variants that replace the constraint violation by a crisp value. From our experiments on both synthetic and real-world benchmark test problems, we find that the performance of the selected algorithms have not been significantly influenced when the constraint violation is not used to guide the environmental selection. Copyright © 2022, The Authors. All rights reserved.

关键词： Benchmarking

A Multi-Layer Long Short-Term Memory Neural Network on Battery State-of-Health Estimation with Transfer Learning

学校读者我要写书评

暂无评论

A Multi-Layer Long Short-Term Memory Neural Network on Batte...

IEEE International Conference on Industrial Electronics for Sustainable Energy Systems (IESES)

作者： Jilun Tian Jiusi Zhang Yuchen Jiang Hao Luo Congsheng Huang Mo-yuen Chow Department of Control Science and Engineering Harbin Institute of Technology Harbin China Department of Electrical Engineering National Taipei University of Technology Taipei Taiwan R.O.C. Department of Electrical and Computer Engineering North Carolina State University Raleigh NC U.S.A.

Due to the important part of batteries in industrial systems, its safety analysis has causes widespread attention from researchers, and its effective maintenance decision-making is needed. Data-driven state-of-health (SOH) estimation can provide useful information by monitoring historical data during the aging process, but it can be failed in the cross-domain scenarios due to the different data distributions. To tackle this issue, we propose a long short-term memory (LSTM) neural network with an additional fully connected dense as a basic predictor, and apply the pretrain and fine-tuning training algorithm to realize the high-performance prediction. By validated in two real-world datasets, we find that blindly expanding the training set may have a negative impact on model accuracy, and the proposed TL-LSTM can achieve a great performance under the cross-domain tasks.

关键词：

Classification of Skin Lesions from Dermatoscopic Images Using Convolutional Neural Networks

学校读者我要写书评

暂无评论

Classification of Skin Lesions from Dermatoscopic Images Usi...

International Conference on control Systems and computer Science (CSCS)

作者： Maria Oniga Alina-Elena Sultana Dan Popescu Daniel-Marian Merezeanu Loretta Ichim Faculty of Automatic Control and Computer Science University Politehnica of Bucharest Bucharest Romania Faculty of Electronics Telecommunications and Information Technology University Politehnica of Bucharest Bucharest Romania “St. S. Nicolau” Institute of Virology Bucharest Romania

Skin cancer is one of the most common types of cancer, and it is caused by a variety of dermatological conditions. Identifying abnormalities from skin images is an important pre-diagnostic step to assist physicians in determining the patient’s condition. Thus, to aid dermatologists in the diagnosis process, we proposed five CNN-based classification approaches namely ResNet-101, DenseNet-121, GoogLeNet, VGG16, and MobileNetV2 architectures on which the transfer learning process was applied. The HAM10000-N database consisting of 7,120 images, which was obtained from the original HAM10000 dataset through an augmentation process, was used to train the proposed methods. Moreover, the images from the HAM10000-N were pre-processed by removing hair with the DullRazor algorithm. To evaluate and compare the performance of all networks five metrics were calculated: accuracy, precision, recall, and Fl-score. The best results for the seven-class classification of the HAM10000-N dataset were obtained for DenseNet-121 architecture with 87% accuracy, 0.871 precision, 0.87 recall and 0.872 F1-score.

关键词：

Shared Situational Awareness with V2X Communication and Set-membership Estimation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Narri, Vandana Alanwar, Amr Mårtensson, Jonas Norén, Christoffer Johansson, Karl Henrik Research and Development Scania CV AB Södertälje151 87 Sweden Division of Decision and Control Systems at School of Electrical Engineering and Computer Science KTH Royal Institute of Technology StockholmSE100 44 Sweden School of Computer Science and Engineering Constructor University Germany

The ability to perceive and comprehend a traffic situation and to estimate the state of the vehicles and road-users in the surrounding of the ego-vehicle is known as situational awareness. Situational awareness for a heavy-duty autonomous vehicle is a critical part of the automation platform and depends on the ego-vehicle’s field-of-view. But when it comes to the urban scenario, the field-of-view of the ego-vehicle is likely to be affected by occlusion and blind spots caused by infrastructure, moving vehicles, and parked vehicles. This paper proposes a framework to improve situational awareness using set-membership estimation and Vehicle-to-Everything (V2X) communication. This framework provides safety guarantees and can adapt to dynamically changing scenarios, and is integrated into an existing complex autonomous platform. A detailed description of the framework implementation and real-time results are illustrated in this paper. Copyright © 2023, The Authors. All rights reserved.

关键词： Vehicle to Everything

A Direct State-Space Realization of Discrete-Time Linear Parameter-Varying Input-Output Models

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Kon, Johan Tóth, Roland van de Wijdeven, Jeroen Heertjes, Marcel Oomen, Tom Control Systems Technology Group Eindhoven University of Technology Netherlands HUN-REN Institute for Computer Science and Control Hungary ASML Veldhoven Netherlands Delft Center for Systems and Control Delft University of Technology Netherlands

A minimal state-space (SS) realization of an identified linear parameter-varying (LPV) input-output (IO) model usually introduces dynamic and nonlinear dependency of the state-space coefficient functions, complicating stability analysis and controller synthesis. The aim of this paper is to introduce and analyze a direct SS realization of this IO model that avoids this nonlinear and dynamic dependency, at the cost of introducing a nonminimal state. It is shown that this direct SS realization 1) is reachable under a coprimeness condition on the coefficient functions of the IO model and a well-posedness condition on the model order, and 2) is never observable but that the unobservable directions converge to zero in a finite amount of steps, i.e., that the realization is reconstructible. The derived results are illustrated through numerical examples in both the LPV and LTI case. © 2025, CC BY-NC-ND.

关键词： State space methods

Machine learning algorithms to predict stroke in China based on causal inference of time series analysis

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zheng, Qizhi Zhao, Ayang Wang, Xinzhu Bai, Yanhong Wang, Zikun Wang, Xiuying Zeng, Xianzhang Dong, Guanghui College of Computer and Control Engineering Northeast Forestry University No.26 Hexing Road Xiangfang District Harbin150040 China Department of Anesthesiology Chongqing University Cancer Hospital No.181 Hanyu Road Shapingba District Chongqing400030 China Harbin Institute of Technology Harbin150001 China

Importance: Identifying and managing high-risk populations for stroke in a targeted manner is a key area of preventive healthcare. Objective: To assess machine learning (ML) models and causal inference of time series analysis for predicting stroke clinically meaningful model. Design: This is a retrospective cohort study and data is from China Health and Retirement Longitudinal Study (CHARLS) assessed 11,789 adults in China from 2011 to 2018. Data analysis was performed from June 1 to December 1, 2024. Setting: CHARLS adopts a multi-stage probability sampling method, covering samples from 28 provinces, and collects data every two years through computer-aided personal interviews (CAPI). Participants: This study employed a combination of Vector Autoregression (VAR) model and Graph Neural Networks (GNN) to systematically construct dynamic causal inference. Multiple classic classification algorithms were compared, including Random Forest, Logistic Regression, XGBoost, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting, and Multi Layer Perceptron (MLP). The SMOTE algorithm was used to undersample a small number of samples and employed Stratified K-fold Cross Validation. Main Outcome(s) and Measure(s): AUC (Area Under the Curve), Accuracy, Precision, Recall, F1 Score, and Matthews Correlation Coefficient (MCC). Results: This study included a total of 11,789 participants, including 6,334 females (53.73%) and 5,455 males (46.27%), with an average age of 65 years. Introduction of dynamic causal inference features has significantly improved the performance of almost all models. The area under the ROC curve of each model ranged from 0.78 to 0.83, indicating significant difference (P © 2025, CC BY.

关键词： Graph neural networks

Approximate Dynamic Programming for Platoon Coordination under Hours-of-Service Regulations

学校读者我要写书评

暂无评论

Approximate Dynamic Programming for Platoon Coordination und...

IEEE Conference on Decision and control

作者： Ting Bai Alexander Johansson Karl Henrik Johansson Jonas Mårtensson Integrated Transport Research Lab and Division of Decision and Control Systems School of Electrical Engineering and Computer Science KTH Royal Institute of Technology Stockholm Sweden Stockholm Sweden

ISBN: (数字)9781665467612

ISBN: (纸本)9781665467629

Truck drivers are required to stop and rest with a certain regularity according to the driving and rest time regulations, also called Hours-of-Service (HoS) regulations. This paper studies the problem of optimally forming platoons when considering realistic HoS regulations. In our problem, trucks have fixed routes in a transportation network and can wait at hubs along their routes to form platoons with others while fulfilling the driving and rest time constraints. We propose a distributed decision-making scheme where each truck controls its waiting times at hubs based on the predicted schedules of others. The decoupling of trucks’ decision-makings contributes to an approximate dynamic programming approach for platoon coordination under HoS regulations. Finally, we perform a simulation over the Swedish road network with one thousand trucks to evaluate the achieved platooning benefits under the HoS regulations in the European Union (EU). The simulation results show that, on average, trucks drive in platoons for 37% of their routes if each truck is allowed to be delayed for 5 % of its total travel time. If trucks are not allowed to be delayed, they drive in platoons for 12 % of their routes.

关键词： Schedules Simulation Roads Decision making Europe Regulation Dynamic programming

Character Recognition of Vedic Sanskrit using Deep Learning Algorithms

学校读者我要写书评

暂无评论

Character Recognition of Vedic Sanskrit using Deep Learning ...

International Conference on Computing, Communication, and Intelligent Systems (ICCCIS)

作者： Mrinal Pandey Rashmikiran Pandey Alexey Nazarov Department of Radio Engineering and Cybernetics Moscow Institute of Physics and Technology Moscow Russia Federal Research Center Computer Science and Control Russian Academy of Sciences Moscow Russia

ISBN: (纸本)9781665462013

One of the applications of deep learning is deciphering the unscripted text over the walls and pillars of historical monuments is the major source of information extraction. This information gives us an idea about the art and the culture of that respective era. In this paper we proposed a model for digital recognition of textual characters written in Sanskrit language using the 4-fold Convolutional Neural Network (CNN) architecture. At the first stage we remove the noise and subsequently segmentation of characters from the input image is performed, later we convert the image into binary image format by the help of image processing techniques. The research is performed over the Devanagari script which was developed during 1 st to the 4 th century CE and is common to many languages from the Indian subcontinent.

关键词： Training Deep learning Support vector machines Computational modeling Natural languages Graphics processing units Information retrieval