November 2024
·
4 Reads
This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.
November 2024
·
4 Reads
November 2024
·
4 Reads
·
1 Citation
Intelligent Systems, IEEE
Advanced communication systems and military reconnaissance are increasingly prevalent in high-tech environments, greatly supported by the flourishing in signal processing technologies. The recent exponential proliferation of sensors led to an unprecedented expansion in the scale and diversity of signals across various modalities. Such influx poses significant challenges in effectively integrating multi-modal signal data to deliver comprehensive and interpretive solutions across a diverse range of applications. In this paper, we provide an overview of the core issues, challenges, and future research directions in different stages of developing large-scale multi-modal signal processing models. Additionally, we introduce a prior investigation into signal representation learning, where we propose a contrastive learning-based framework to extract fine-grained signal features under few-shot conditions. Our proposed framework achieves a 24.1% performance improvement over baseline approaches, consistently demonstrating superiority over state-of-the-art methods. The code is accessible in this repository: https://github.com/YYH211/LSM .
January 2024
·
28 Reads
IEEE Network
Streaming video analytics focuses on the real-time analysis of streaming video data from multiple resources, such as security cameras, and IoT devices with video capabilities. It involves applications of various techniques to extract valuable information from live video streams. Edge computing and cloud computing facilitate video stream analytics by utilizing computation resources across both ends, enabling both high accuracy and low latency. However, video streaming behaviours are dynamic and constantly evolving across the edge and the cloud. The network conditions, computing resources, and video content can change rapidly, making it crucial to continuously adjust the analytics methods to provide accurate results. Previous works both based on deep neural networks (DNNs) or heuristic algorithms learn a suitable deployment plan for streaming video analytics applications from historical data or synthetic data and therefore are not able to capture the dynamics. Hence, we propose reinforcement learning-based methods that can adapt to ongoing changes in video streaming behaviours. To ensure the scalability of video analytics in distributed environments, we implement OSMOTICGATE2, a distributed streaming video analytics system that features optimized processing pipelines and multi-agent RL-based controllers for fast adapting the system configurations across the edge and the cloud. Experiments on a real testbed show that our method outperforms baselines, assuring real-time video analysis and high accuracy in dynamic and distributed environments.
January 2024
·
6 Reads
IEEE Transactions on Industrial Cyber-Physical Systems
Detecting defects on railway tracks is critical for the operation of high-speed trains. Despite a plethora of machine vision-based methods designed to tackle this problem, the majority adopt a supervised setting and demand considerable labeled training data, inclusive of defective samples, which is expensive and impractical. In this paper, we propose an I nvertible R econstruction neural N etwork (IRNet) for semi-supervised rail surface defect detection, where only normal images are accessible during training. Firstly, we devise an information-preserving feature encoder comprising several invertible blocks. This structure safeguards subtle visual patterns distinguishing normal and defective images from being obscured by background information, guaranteed by its mathematical reversibility property. Second, to overcome the overgeneralization issue of conventional autoencoders caused by imperfectly crafted decoders, we propose a novel decoder-free reconstruction workflow based on the invertible feature encoder. Specifically, we force one portion of extracted features to approach a predefined constant tensor during the training stage by minimizing their mean squared error. Next, we feed the remained features and the predefined constant tensor backward into the encoder to reconstruct the original images. During the testing phase, we formulate an anomaly score that consolidates the reconstruction error and mean squared error to spot defects. Extensive experiments are conducted on 4 real-world datasets. Our method consistently outperforms state-of-the-art techniques, demonstrating an average increase of 8.5% on the F1 score.
December 2023
·
47 Reads
·
69 Citations
IEEE Transactions on Wireless Communications
Edge caching is a promising approach to reduce duplicate content transmission in Internet-of-Vehicles (IoVs). Several Reinforcement Learning (RL) based edge caching methods have been proposed to improve the resource utilization and reduce the backhaul traffic load. However, they only obtain the local sub-optimal solution, as they neglect the influence from environments by other agents. This paper investigates the edge caching strategies with consideration of the content delivery and cache replacement by exploiting the distributed Multi-Agent Reinforcement Learning (MARL). A hierarchical edge caching architecture for IoVs is proposed and the corresponding problem is formulated with the goal to minimize the long-term content access cost in the system. Then, we extend the Markov Decision Process (MDP) in the single agent RL to the context of a multi-agent system, and tackle the corresponding combinatorial multi-armed bandit problem based on the framework of a stochastic game. Specifically, we firstly propose a Distributed MARL-based Edge caching method (DMRE), where each agent can adaptively learn its best behaviour in conjunction with other agents for intelligent caching. Meanwhile, we attempt to reduce the computation complexity of DMRE by parameter approximation, which legitimately simplifies the training targets. However, DMRE is enabled to represent and update the parameter by creating a lookup table, essentially a tabular-based method, which generally performs inefficiently in large-scale scenarios. To circumvent the issue and make more expressive parametric models, we incorporate the technical advantage of the Deep- Q Network into DMRE, and further develop a computationally efficient method (DeepDMRE) with neural network-based Nash equilibria approximation. Extensive simulations are conducted to verify the effectiveness of the proposed methods. Especially, DeepDMRE outperforms DMRE, Q -learning, LFU, and LRU, and the edge hit rate is improved by roughly 5%, 19%, 40%, and 35%, respectively, when the cache capacity reaches 1, 000 MB.
October 2023
·
96 Reads
·
1 Citation
Mathematics
The arrival interval at high-speed railway stations is one of the key factors that restrict the improvement of the train following intervals. In the process of practical railway operation, sudden conflicts occur sometimes. Especially when the conflict arises at the station, because the home signal cannot be opened in time, the emergency may affect the adjustment of the train operation under the scheduled timetable, resulting in a longer train following interval or even delay. With the development of artificial intelligence and the deep integration of big data, the architecture of train operation control and dispatch integration is gradually improving from the theoretical point. Based on this and inspired by the Green Wave policy, we propose an integrated operation method that reduces the arrival interval by avoiding unnecessary stops in front of the home signal and increasing the running speed of trains through the throat area. It is a two-step optimization method combining both intelligent optimization and mathematical–theoretical analysis algorithms. In the first step, the recommended approaching speed and position are obtained by analytical calculation. In the second step, the speed profile from the current position to the position corresponding to the recommended approaching speed is optimized by intelligent optimization algorithms. Finally, the integrated method is verified through the analysis of two distinct case studies. The first case study utilizes data from the Beijing–Shanghai high-speed railway line, while the second one is based on the field test. The numerical result shows that the proposed method could save the entry running time effectively, compared with the normal strategy given by the train driver. The method can mitigate controllable conflict events occurring at the station and provides theoretical support for practical operation.
May 2023
·
27 Reads
·
82 Citations
IEEE Transactions on Network Science and Engineering
This paper considers computation offloading and service caching in a three-tier mobile cloud-edge computing structure, in which Mobile Users (MUs) have subscribed to the Cloud Service Center (CSC) for computation offloading services and paid related fees monthly or yearly, and the CSC provides computation services to subscribed MUs and charges service fees. Long transmission distance and communication resource shortage caused by the increasing number of offloaded MUs may make the CSC unable to satisfy the delay requirements of MUs. Hence, the CSC can purchase some computation and communication resources from Edge Servers (ESs) with limited caching capacities and computation resources to assist MUs in computation offloading. However, from the perspective of the CSC, it remains open to jointly optimize the strategies of computation offloading, service caching, and resource allocation to meet the delay requirements of MUs while reducing the cost of the CSC. Therefore, a novel Deep Reinforcement Learning-based Computation Offloading and Service Caching Mechanism, named DRLCOSCM is proposed to jointly optimize the offloading decision, service caching, and resource allocation strategies, so as to minimize the cost of the CSC while ensuring the delay requirements of MUs. In DRLCOSCM, the optimization problem is formulated as a Mixed Integer Non-Linear Programming (MINLP) problem, and an Asynchronous Advantage Actor-Critic-based (A3C-based) algorithm is proposed to solve the problem. The simulation results show that DRLCOSCM significantly outperforms the other baseline methods in different scenarios.
February 2023
·
57 Reads
·
58 Citations
IEEE Robotics and Automation Letters
Implicit neural representations have shown compelling results in offline 3D reconstruction and also recently demonstrated the potential for online SLAM systems. However, applying them to autonomous 3D reconstruction, where a robot is required to explore a scene and plan a view path for the reconstruction, has not been studied. In this paper, we explore for the first time the possibility of using implicit neural representations for autonomous 3D scene reconstruction by addressing two key challenges: 1) seeking a criterion to measure the quality of the candidate viewpoints for the view planning based on the new representations, and 2) learning the criterion from data that can generalize to different scenes instead of a hand-crafting one. To solve the challenges, firstly, a proxy of Peak Signal-to-Noise Ratio (PSNR) is proposed to quantify a viewpoint quality; secondly, the proxy is optimized jointly with the parameters of an implicit neural network for the scene. With the proposed view quality criterion from neural networks (termed as Neural Uncertainty), we can then apply implicit representations to autonomous 3D reconstruction. Our method demonstrates significant improvements on various metrics for the rendered image quality and the geometry quality of the reconstructed 3D models when compared with variants using TSDF or reconstruction without view planning. Project webpage https://kingteeloki-ran.github.io/NeurAR/</uri
January 2023
·
46 Reads
·
2 Citations
January 2023
... [38]- [43] into SLAM systems, eliminating the need for handcrafted feature extraction by optimizing end-to-end loss functions to learn the required features and representations from input data. Compared to traditional SLAM methods, learning-based SLAM systems typically exhibit better accuracy and robustness. ...
November 2024
Intelligent Systems, IEEE
... Zhong et al. [26] proposed a service caching algorithm based on multi-agent actor-critic deep reinforcement learning to minimize system transmission delay when the popularity of user preferences is unknown. Zhou et al. [27] studied the service deployment strategy considering service distribution and replacement, and proposed a reinforcement learning algorithm based on multi-agent collaboration to minimize the cost of service, in which each edge server makes the best local decision. Yao et al. [28] studied the joint problem of task scheduling and service caching based on the cooperation of edge servers, and proposed a MARL algorithm combined with graph attention network to learn the joint strategy. ...
December 2023
IEEE Transactions on Wireless Communications
... [42] proposed a joint computing offloading and content caching scheme called CoPace. In this, OSTP is introduced for predicting the popularity of future time-varying con- Latency [32] Latency [33] Latency& Energy [40] Latency [41] Latency& Energy [42] Latency [43] Latency& Cost ...
May 2023
IEEE Transactions on Network Science and Engineering
... Iversen et al. [23] exploit a noise model [5] for simulating Kinect V1 depth images, which can be used in various computer vision algorithms requiring a substantial dataset. Recently, Ran et al. [30] have introduced an implicit neural representation for path planning and active reconstruction, using the sensor noise model to assess reconstruction uncertainty to determine the next best view for active scanning. Katar et al. [31] model lateral and axial noise in 3D camera data to enhance synthetic training datasets, demonstrating that this improves neural network performance in real-world object segmentation tasks. ...
February 2023
IEEE Robotics and Automation Letters
... Jie et al. [22] proposed a depth-first search crew recovery (DFSCR) method to deal with the real-time crew rescheduling problem. Liu et al. [23] proposed a real-time rescheduling model combined ATO driving strategy to restore the train operation from the delay caused by disturbance. Zhang et al. [24] developed an efficient heuristic algorithm to solve the train rescheduling problem in a railway network with the goal of reducing passenger inconvenience. ...
October 2022
... This indicates that as the network grows, it adapts to have the scalefree property to gain robustness against failures. A recent study of the brain network of C. elegans across its development found that degree distributions consistently did not follow a power-law model across all stages [49]. This observation underscores the intricacies inherent in the developmental process of Drosophila compared to C. elegans, a distinction that aligns with the fact that the former undergoes complete metamorphosis. ...
December 2022
Entropy
... Most of them were developed before significant progress was made in single-cell atlas, and the cost of bulk-level data is less than sequencing single-cell data [25]. Nvwa [26] attempted to predict whether a gene will express in cells using single-cell transcriptomics, but non-human data were considered, and classification is not a proper problem setting for gene expression prediction because it is important to evaluate the expression levels. Recently, seq2cells [11] built on the pre-trained Enformer model to extract embeddings from DNA sequences to predict gene expression across all the cells, whereas scooby [27] used a similar idea for single-cell multi-omic data. ...
October 2022
Nature Genetics
... Karim et al. discussed the objective QoE assessment of original video quality and compressed videos from Face book, Qzone, and Tumblr [8]. Even though, the mobile devices are resource constrained, the mobile applications require QoS parameters to be satisfied [9]. For this purpose, we Content courtesy of Springer Nature, terms of use apply. ...
January 2022
IEEE Transactions on Mobile Computing
... Recently, deep learning-based detection methods have advanced considerably, allowing for adaptive feature extraction based on image characteristics and showing significant potential for applications in railways [27]. References [28] and [29] have developed dual-branch networks with cross-linking based on deep learning to detect railway pedestrians. However, these methods face significant challenges when applied to real world railway scenarios with complex and variable backgrounds. ...
January 2022
IEEE Transactions on Intelligent Vehicles
... For example, cell atlases for vertebrate systems covering fetal and adult periods have been generated, such as Xenopus (12), zebrafish embryos (13)(14)(15), Human Cell Landscape (HCL) (16), Tabula Sapiens (17), Mouse Cell Atlas (MCA) (8), Mouse Cell Differentiation Atlas (MCDA) (18), Tabula Muris (19), Tabula Muris Senis (20) and Zebrafish Cell Landscape (ZCL) (21). Several invertebrate cell atlases are also available, including for Caenorhaditis elegans (22), Drosophila embryo (11), Nematostella (23), sea squirt (24), fruitfly (25) and earthworm (26). These studies have significantly enriched our knowledge about cellular hierarchy in different species. ...
April 2022