Article
Engineering, Electrical & Electronic
Pengwen Xiong, Xiaobao Tong, Aiguo Song, Peter X. Liu
Summary: This study proposes a novel adaptive multikernel dictionary learning method for analyzing the grasping tactile information of mechanical dexterous grippers. By using multiple basic kernel functions and an adaptive kernel weight calculation method, the force coupling among fingers is effectively considered by combining the tactile information from multiple fingers.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Computer Science, Artificial Intelligence
Yuanjie Zheng, Yunshuai Yang, Tongtong Che, Sujuan Hou, Wenhui Huang, Yue Gao, Ping Tan
Summary: This research observes a common characteristic between classical image matting and Gaussian process-based regression and proposes a new formulation of image matting using GP. The use of kernel learning in GP leads to the development of a powerful deep matting-GP technique with reduced computational complexity. Experimental results demonstrate the superiority of deep matting-GP over traditional matting techniques and modern deep learning approaches on both synthetic and real-world images.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Interdisciplinary Applications
J-L Akian, L. Bonnet, H. Owhadi, E. Savin
Summary: This paper introduces algorithms for selecting/designing kernels in Gaussian process regression/kriging surrogate modeling techniques. It presents two classes of algorithms: kernel flow, which selects the best kernel by minimizing the loss of accuracy caused by removing a portion of the dataset, and spectral kernel ridge regression, which selects the best kernel by minimizing the norm of the function to be approximated in the associated RKHS. The effectiveness of both approaches is demonstrated through numerical examples.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)
Article
Engineering, Civil
Zhiyuan Liu, Cheng Lyu, Jinbiao Huo, Shuaian Wang, Jun Chen
Summary: Gaussian process regression (GPR) is a promising machine learning model for transportation system estimation and prediction. The radial basis function (RBF) kernel, commonly used in GPR, often faces difficulties in finding the optimal hyperparameter. This paper addresses the issue by promoting the use of the hat kernel and investigating the connection between deformation and the Bayesian generalization error of GPR.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Peng Wang, Lyudmila Mihaylova, Said Munir, Rohit Chakraborty, Jikai Wang, Martin Mayfield, Khan Alam, Muhammad Fahim Khokhar, Daniel Coca
Summary: This paper introduces a computationally efficient Gaussian process approach that achieves better computational efficiency compared with standard methods when using fewer data. The approach incorporates the 'residual' matrix in its symmetric diagonally dominant form and can further approximate it by the Neumann series.
Article
Chemistry, Multidisciplinary
Hongchang Liu, Mingfang He, Weiwei Cai, Guoxiong Zhou, Yanfeng Wang, Liujun Li
Summary: This study proposes a working condition recognition method for mineral flotation process based on deep and shallow feature fusion. By fusing the extracted shallow features and deep features, the method eliminates redundant and noisy information and achieves effective recognition of the working condition.
APPLIED SCIENCES-BASEL
(2022)
Article
Robotics
Weizhe Chen, Roni Khardon, Lantao Liu
Summary: Robotic Information Gathering (RIG) is a foundational research topic that addresses the efficient collection of informative data to build an accurate model of an unknown target function under robot embodiment constraints. RIG has many applications, including autonomous exploration and mapping, 3D reconstruction or inspection, search and rescue, and environmental monitoring. RIG systems rely on probabilistic models and prediction uncertainty to identify critical areas for data collection.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Pengfei Wei, Thanh Vinh Vo, Xinghua Qu, Yew Soon Ong, Zejun Ma
Summary: In this article, an effective method of explicitly modeling the domain relatedness of each domain pair through transfer kernel learning is proposed. To overcome the limitations of existing transfer kernels, a novel multi-source transfer kernel k(ms) is further introduced. The proposed method assigns a learnable parametric coefficient to model the relatedness of each inter-domain pair and simultaneously regulates the relatedness of the intra-domain pair to be 1. Experimental results demonstrate the effectiveness of the proposed method in domain relatedness modeling and transfer performance.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Lixiang Xu, Biao Zhou, Xinlu Li, Zhize Wu, Yan Chen, Xiaofeng Wang, Yuanyan Tang
Summary: This paper proposes an image classification method based on the average weight selective kernel network (AWSKnet) model. By incorporating the idea of ensemble learning and leveraging convolution layer features, the method improves the effectiveness of feature training. Additionally, a multi-layer convolution kernel is constructed using a base kernel function, and the method outperforms state-of-the-art image classification models, as demonstrated by experimental results.
Article
Computer Science, Artificial Intelligence
Tingting Wang, Lei Xu, Junbao Li
Summary: Deep convolutional neural networks have great potential in image recognition tasks, but are hindered by their complexity and potential overfitting. Kernel learning methods offer a clear mathematical theory and fewer parameters, but struggle with high-dimensional data. The proposed SDCRKL-GP method effectively combines deep convolutional architecture with kernel learning, achieving excellent performance and low complexity, outperforming state-of-the-art algorithms in accuracy and speed for image recognition tasks.
Article
Computer Science, Artificial Intelligence
Jiyang Xie, Zhanyu Ma, Dongliang Chang, Guoqiang Zhang, Jun Guo
Summary: Channel attention mechanisms are widely used in visual tasks for performance improvement. The GPCA module proposed in this paper models correlations among channels using a Gaussian process and solves mathematical tractability issues with a Sigmoid-Gaussian approximation.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Multidisciplinary Sciences
Adel Parvizi-Fard, Mahmood Amiri, Deepesh Kumar, Mark M. Iskarous, Nitish Thakor
Summary: The study modeled three stages of the tactile pathway from the periphery to the cortex and explored the roles of sensory neurons, cuneate nucleus, and cortical neurons in sensing indentation and edge stimuli. By utilizing a biomimetic decoder, edge orientation can be detected more accurately, and it was observed that larger receptive fields convey more information about edge orientation.
SCIENTIFIC REPORTS
(2021)
Article
Automation & Control Systems
Tianliang Li, Jinxiu Guo, Han Zheng, Shasha Wang, Liang Qiu, Hongliang Ren
Summary: A six-axis force/moment tactile sensor based on fiber Bragg grating (FBG) has been developed for robot-assisted minimally invasive surgery, with the capabilities of nonlinear decoupling, fault-tolerant, and temperature compensation. The sensor consists of eight tightly suspended FBGs fixed inside a small three-dimensional printed flexure. An optimized back propagation neural network (BPNN) algorithm has been proposed to eliminate nonlinear crosstalk, diagnose and modify fault response, and eliminate temperature-induced errors.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)
Article
Automation & Control Systems
Amit Kumar Naik, Prabhat Kumar Upadhyay, Abhinoy Kumar Singh
Summary: The solution to practical nonlinear filtering problems relies on Gaussian filtering, which involves intractable integrals that are numerically approximated. This paper proposes a new quadrature rule based Gaussian filter, named Gaussian kernel quadrature Kalman filter (GKQKF), which improves the numerical approximation accuracy and estimation accuracy compared to existing Gaussian filters.
EUROPEAN JOURNAL OF CONTROL
(2023)
Article
Energy & Fuels
Jianjun Chen, Weihao Hu, Di Cao, Man Zhang, Qi Huang, Zhe Chen, Frede Blaabjerg
Summary: Traditional AI-based fault detection methods require extensive labeled data which is difficult to obtain. A novel few-shot learning method based on deep Gaussian process kernel transfer is proposed for accurate fault detection in electric machines under variable working conditions, outperforming conventional and other few-shot learning methods.
IEEE TRANSACTIONS ON ENERGY CONVERSION
(2021)
Review
Computer Science, Hardware & Architecture
Martin Maier, Amin Ebrahimzadeh, Abdeljalil Beniiche, Sajjad Rostami
Summary: The paper discusses the factors driving network evolution and the complexity of modern networks, exploring the art of future 6G mobile networks, focusing on optical fiber fixed networks and digital society, presenting human-centric Industry 5.0 and its principles, and introducing a token engineering framework designed for Society 5.0 based on CPSS.
JOURNAL OF OPTICAL COMMUNICATIONS AND NETWORKING
(2022)
Article
Engineering, Multidisciplinary
Marsa Rayani, Amin Ebrahimzadeh, Roch H. Glitho, Halima Elbiaze
Summary: This paper proposes a framework for content delivery using information-centric networks and IP slices. By leveraging in-network caching advantages, it addresses the creation of ICN and IP slices in a dynamic environment and achieves cost-efficient and profit-maximizing solutions through a dynamic network slicing heuristic algorithm.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Information Systems
Mouhamad Dieye, Amina Mseddi, Wael Jaafar, Halima Elbiaze
Summary: This paper proposes a reliable fog-based remote health monitoring framework operating under uncertain fog computing conditions. By using reinforcement learning and differential evolution-based algorithm, the framework maximizes the number of satisfied tasks and demonstrates superior reliability performance compared to benchmarks.
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT
(2022)
Article
Computer Science, Information Systems
Cheska C. Abarro, Angela C. Caliwag, Erick C. Valverde, Wansu Lim, Martin Maier
Summary: This article proposes an IoT-based low-delay smart streetlight monitoring system that provides real-time monitoring and greatly reduces data storage usage through a data filtering algorithm.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Computer Science, Information Systems
Sami Nadif, Essaid Sabir, Halima Elbiaze, Abdelkrim Haqiq
Summary: This article presents the challenges of using narrowband Internet of Things (NB-IoT) technology in an ultradense small cell network and proposes a power allocation method to address these challenges. By leveraging stochastic geometry analysis and mean-field game theory, a consistent and distributed solution is provided, and its effectiveness is demonstrated through numerical analysis.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Fatemeh Aghaaliakbari, Zakaria Ait Hmitti, Marsa Rayani, Manel Gherari, Roch H. Glitho, Halima Elbiaze, Wessam Ajib
Summary: This article advocates the use of in-network computing (INC) paradigm to tackle the high bandwidth and low latency challenges of holographic applications, instead of the previously used edge computing paradigm. An architecture is proposed for provisioning INC-enabled slices for holographic-type application deployment, which is validated through a proof of concept and extensive simulations. Experimental results show that INC outperforms edge computing in addressing these key challenges, while maintaining low jitter for hologram stability.
IEEE COMMUNICATIONS MAGAZINE
(2023)
Article
Computer Science, Information Systems
Amina Mseddi, Wael Jaafar, Halima Elbiaze, Wessam Ajib
Summary: Fog computing is a new paradigm for enabling IoT applications. Due to the complexity of fog infrastructure, resource coordination and tracking are challenging. In this study, we propose online resource allocation solutions to maximize user satisfaction within latency requirements, using Markov Decision Process and reinforcement learning. Simulation results based on real-world data demonstrate the high efficiency of our collaborative solution compared to existing methods.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Proceedings Paper
Computer Science, Information Systems
Muhammad Saqib, Zakaria Ait Hmitti, Halima Elbiaze, Roch H. Glitho
Summary: In-network traffic classification is a category of in-network computing, where an accurate network traffic classifier is deployed inside a programmable data plane to classify traffic at maximum speed while considering device constraints. Traffic flow can be classified using a single feature, sequential packet size information, to achieve accurate and early-stage network traffic classification with minimal use of networking device resources.
2022 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM 2022)
(2022)
Proceedings Paper
Computer Science, Information Systems
Aya Ahmed, Cirine Chaieb, Wessam Ajib, Halima Elbiaze, Roch Glitho
Summary: This paper investigates the use of cell-free unmanned aerial vehicles (UAVs)-assisted wireless networks and optimizes the number of deployed UAVs under quality of service and coverage constraints. The proposed low-complexity efficient greedy-based algorithmic solutions can tackle the user-UAVs association, UAVs placement, channel assignment, and transmit power allocation. Simulation results illustrate the efficiency of the proposed algorithms in terms of the number of deployed UAVs in cell-free wireless networks.
2022 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM 2022)
(2022)
Proceedings Paper
Computer Science, Information Systems
Soukaina Ouledsidi Ali, Halima Elbiaze, Roch Glitho, Wessam Ajib
Summary: Microservices are a promising technology for future networks, but their deployment in edge and in-network devices is more expensive and requires consideration of specific requirements. This paper proposes a heuristic solution to the problem of microservices placement, demonstrating its superiority in cost reduction and latency minimization.
2022 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM 2022)
(2022)
Proceedings Paper
Computer Science, Information Systems
Sami Nadif, Essaid Sabir, Halima Elbiaze, Oussama Habachi, Abdelkrim Haqiq
Summary: This paper investigates the location and power allocation problem for enhanced Mobile Broadband (eMBB) users and Internet of Things (IoT) devices. It proposes a hierarchical mean-field game model using the Stackelberg-Nash differential game and mean-field approximation methods, and demonstrates the optimal power allocation strategies.
2022 20TH INTERNATIONAL SYMPOSIUM ON MODELING AND OPTIMIZATION IN MOBILE, AD HOC, AND WIRELESS NETWORKS (WIOPT 2022)
(2022)
Proceedings Paper
Computer Science, Hardware & Architecture
Fatemeh Aghaaliakbari, Farzaneh Ghasemi Javid, Zarin Tasnim, Zakaria Ait Hmitti, Manel Gherari, Roch H. Glitho, Halima Elbiaze
Summary: Holographic applications have demanding requirements that current network infrastructure struggles to meet. The in-network computing paradigm offers a promising solution by distributing computing workloads across the network, resulting in improved low latency and high bandwidth capabilities.
2022 IEEE CONFERENCE ON NETWORK FUNCTION VIRTUALIZATION AND SOFTWARE DEFINED NETWORKS (IEEE NFV-SDN)
(2022)
Proceedings Paper
Computer Science, Hardware & Architecture
Zakaria Ait Hmitti, Hamza Ben Ammar, Ece Gelal Soyak, Youcef Kardjadja, Sepideh Malektaji, Soukaina Ouledsidi Ali, Marsa Rayani, Muhammad Saqib, Seyedreza Taghizadeh, Wessam Ajib, Halima Elbiaze, Ozgur Ercetin, Yacine Ghamri-Doudane, Roch Glitho
Summary: This paper proposes a vision for reshaping the current network infrastructure towards a Next-Generation Networking Infrastructure (NGNI) by combining collaborative computing, caching and communication paradigm with artificial intelligence. This aims to fulfill the stringent requirements of emerging applications.
2022 31ST INTERNATIONAL CONFERENCE ON COMPUTER COMMUNICATIONS AND NETWORKS (ICCCN 2022)
(2022)
Proceedings Paper
Computer Science, Hardware & Architecture
Sami Nadif, Essaid Sabir, Halima Elbiaze, Abdelkrim Haqiq
Summary: This paper explores grant-free access with multi-packet reception capabilities, focusing on ultra-low-end IoT applications with small data sizes and energy usage constraints. It proposes a power allocation scheme that integrates the IoT device's traffic and energy budget using stochastic geometry framework and mean-field approximation. It also derives a Markov model to capture the IoT device's queue length and the successful transmission probability at steady state.
2022 20TH MEDITERRANEAN COMMUNICATION AND COMPUTER NETWORKING CONFERENCE (MEDCOMNET)
(2022)
Article
Telecommunications
Vahid Maleki Raee, Amin Ebrahimzadeh, Roch H. Glitho, Halima Elbiaze
Summary: Traditional wireless sensor networks have high deployment and maintenance costs due to embedded applications. Virtualization technologies address this issue but come with energy-delay cost. This research aims to solve the task assignment challenge in virtualized WSNs while minimizing energy consumption and meeting deadlines.
IEEE TRANSACTIONS ON GREEN COMMUNICATIONS AND NETWORKING
(2022)
