Searching for key nodes and edges in a network is long-standing problem. Recently cycle structure has received more attention. Is it possible to propose ranking algorithm importance? We address the problem of identifying cycles network. First, we provide concrete definition importance-in terms Fiedler value (the second smallest Laplacian eigenvalue). Key are those that contribute most substantially dynamical behavior Second, by comparing sensitivity different cycles, neat index provided....

Federated Learning (FL) is an emerging learning paradigm that enables the collaborative of different nodes without ex-posing raw data. However, a critical challenge faced by current federated algorithms in real-world applications long-tailed data distribution, i.e., both local and global views, numbers classes are often highly imbalanced. This would lead to poor model accuracy on some rare but vital classes, e.g., those related safety health autonomous driving applications. In this paper, we...

Synchronizability is characterized by the smallest real part of nonzero Laplacian eigenvalues, known as algebraic connectivity. It that adding edges an effective strategy for improving network synchronizability in undirected networks. A natural question is, does this conclusion also hold directed networks? This paper aims to answer question. Utilizing eigenvectors, approach focusing on accelerating synchronization networks through edge-addition or edge-removal proposed providing approximate...

Many complex multi-agent systems such as robot swarms control and autonomous vehicle coordination can be modeled Multi-Agent Reinforcement Learning (MARL) tasks. QMIX, a widely popular MARL algorithm, has been used baseline for the benchmark environments, e.g., Starcraft Challenge (SMAC), Difficulty-Enhanced Predator-Prey (DEPP). Recent variants of QMIX target relaxing monotonicity constraint allowing performance improvement in SMAC. In this paper, we investigate code-level optimizations...

Fiedler value, as the minimal real part of (or minimal) nonzero Laplacian eigenvalue, garners significant attention a metric for evaluating network topology and its dynamics. In this paper, we address quantification relation between value each edge in directed complex network, considering undirected networks special case. We propose an approach to measure dynamical contribution edge. Interestingly, these values can be both positive negative, which are determined by left right vectors....

Reinforcement learning (RL) has achieved impressive performance in various domains. However, most RL frameworks oversimplify the problem by assuming a fixed-yet-known environment and often have difficulty being generalized to real-world scenarios. In this paper, we address new challenge with more realistic setting, Incremental Learning, where search space of Markov Decision Process continually expands. While previous methods usually suffer from lack efficiency exploring unseen transitions,...

In this study, to explore the demand characteristics of autonomous coaches in mixed traffic flow, two sections an expressway were selected for vehicle experiments. With personification as goal, sensors ego-vehicle used collect naturalistic car-following behavior data surrounding coaches. After analyzing data, car-driving coach drivers acquired. The analysis results indicate that overall processes tend be relatively stable, and most are conditions higher velocity, smaller acceleration,...

This paper proposes a method for visibility detection based on the recognition of preceding vehicle's taillight signals using in-vehicle camera and millimeter-wave (mm-W) radar. First, we design two methods vehicle identification. One is to use Haar-like features an AdaBoost algorithm train classifier, which mainly used identify vehicles without turning taillights. The other with taillights by means segmentation. identification are combined Discriminative Scale Space Tracker (DSST) track in...

Internet of Things (IoT) interconnects a massive amount devices, generating heterogeneous data with diverse characteristics. IoT emerges as vital asset for data-intensive applications, such healthcare, smart city and predictive maintenance, harnessing the vast volume to its maximum advantage. These applications leverage different Artificial Intelligence (AI) algorithms discover new insights. While machine learning effectively uncovers implicit patterns through model training, centralizing...

Recently, deep neural networks have achieved remarkable progress in class balancing instance segmentation. However, most applications the real world a long-tailed distribution, i.e., limited training examples majority of classes. The challenge leads to catastrophic drop segmentation because gradient head classes suppresses tail classes, leading bias towards major We propose LiCAM, novel framework for It features an adaptive loss function named Moac Loss, which is adjustable during according...

Tall and skinny QR (TSQR) decomposition is an essential matrix operation with various applications in edge computing, including data compression, subspace projection, dimension reduction. As a critical component TSQR, Dual-Triangular (DTQR) solved by the Normal method most works without utilizing dual-triangular structure. Therefore, we propose novel DTQR accelerator recursively exploring DT structure three acceleration strategies systolic array to achieve higher parallelism. Experimental...

Due to improve the rate of accuracy face detection when illuminating condition, position, expression was changeable, proposed LTP (local ternary patterns) features image which is a generalization local binary patterns (LBP) and its texture descriptor more discriminate less sensitive noise. Because feature very robust illustration change rotation face, it could be used solve low recognition that Haar-like method detects under un-normal conditions. It experimented on large difficult...

Large language models (LLMs) have the potential to transform digital healthcare, as evidenced by recent advances in LLM-based virtual doctors. However, current approaches rely on patient's subjective descriptions of symptoms, causing increased misdiagnosis. Recognizing value daily data from smart devices, we introduce a novel multi-turn consultation doctor system, DrHouse, which incorporates three significant contributions: 1) It utilizes sensor devices diagnosis process, enhancing accuracy...

In sectors such as finance and healthcare, where data governance is subject to rigorous regulatory requirements, the exchange utilization of are particularly challenging. Federated Learning (FL) has risen a pioneering distributed machine learning paradigm that enables collaborative model training across multiple institutions while maintaining decentralization. Despite its advantages, FL vulnerable adversarial threats, poisoning attacks during aggregation, process typically managed by central...

<title>Abstract</title> In sectors such as finance and healthcare, where data governance is subject to rigorous regulatory requirements, the exchange utilization of are particularly challenging. Federated Learning (FL) has risen a pioneering distributed machine learning paradigm that enables collaborative model training across multiple institutions while maintaining decentralization. This approach inherently heightens privacy by sharing only weights, rather than raw data. Despite its...