Dear editor,Book thickness for graphs forms a major theme in graph theory and has a broad application in the fields of sorting permutations, fault tolerant VLSI design, parallel computing, and others. The book Thickne...
详细信息
Dear editor,Book thickness for graphs forms a major theme in graph theory and has a broad application in the fields of sorting permutations, fault tolerant VLSI design, parallel computing, and others. The book Thickness problem, even if the vertex order is taken as part of the input, was shown to be NP-complete in general [1, 2]. Just because of this, parameterized algorithms have been proposed to deal with it [3].
Continual Semantic Segmentation (CSS) aims to continuously learn new classes while mitigating catastrophic forgetting. Existing CSS methods primarily address this challenge through knowledge distillation. While they f...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Continual Semantic Segmentation (CSS) aims to continuously learn new classes while mitigating catastrophic forgetting. Existing CSS methods primarily address this challenge through knowledge distillation. While they focus on maintaining stability for old classes, this emphasis often restricts plasticity in learning new ones. In this work, we propose two plug-and-play methods to better balance stability and plasticity. First, for feature map distillation, we replace pixel-level constraints with statistical information extraction within a window, improving the learning of new classes while preserving structural integrity to overcome forgetting. Second, we use relative ranks of predictions to replace exact probabilistic values to loosen the constraints in the logit distribution of old classes. Extensive experiments demonstrate that our methods outperform state-of-the-art methods when integrating into several baselines across various CSS scenarios.
This paper introduced a dual-layer broadband high-gain circularly polarized (CP) antenna. The proposed broadband CP antenna consists of the radiating patch and couping structure, which is made up of opening slot patch...
详细信息
To achieve progressive and accurate decision-making for long-term time series data while meeting the needs of privacy-friendly and early, this paper proposes a universal framework for sequential progressive decision-m...
To achieve progressive and accurate decision-making for long-term time series data while meeting the needs of privacy-friendly and early, this paper proposes a universal framework for sequential progressive decision-making (SPD). This framework first segments the data and sets up multiple columns of neural networks according to the number of segments. Each column can make segmented decisions based on the inputs for the period. Additionally, without sharing the original data, the framework leverages lateral hidden layer connections between preceding and succeeding columns to obtain useful features for subsequent decision-making, gradually improving accuracy while avoiding the risk of data leakage. SPD has the advantages of privacy friendliness, column model diversification, prior knowledge reuse, and easy scalability, making it an effective framework for continuous decision-making. The effectiveness of that was validated using various network models in handwritten digit recognition and electrocardiogram classification tasks. The obtained experimental results reveal that SPD not only enables early decision-making while ensuring accuracy but also achieves accuracy levels comparable to or even surpassing those obtained using complete data, with the added benefit of privacy protection.
How to schedule public resources to maximize the coverage of crowds has always been a hot topic. There are mainly two types of research. One is the scheduling of static resources, whose coverage efficiency is lower be...
How to schedule public resources to maximize the coverage of crowds has always been a hot topic. There are mainly two types of research. One is the scheduling of static resources, whose coverage efficiency is lower because the resources cannot be moved; the other is the dynamic scheduling of movable public resources, but the existing methods require additional devices to collect crowd volume data. To address this issue, this article introduces a new category of mobile agents that can sense and execute simultaneously. This means that an agent can simultaneously perform crowd volume data collection tasks and crowds coverage tasks in its area. Taking these multi-ability agents as public resources, this article proposes a scheduling method called GE-STMC (Greedy Execute - Spatial and Temporal Monte-Carlo dropout). The method considers the spatiotemporal correlation between areas and evaluates the sensing value and the executing value of areas comprehensively to allocate the agents accordingly. Through the experimental on the real data of Beijing Happy Valley, this method can get 16% larger crowds volume coverage than other baseline methods with the same number of agents.
Global routing is a crucial step in the design of Very Large-Scale Integration (VLSI) circuits. However, most of the existing methods are heuristic algorithms, which cannot conjointly optimize the subproblems of globa...
详细信息
In wireless communication systems, accurate channel estimation is essential to ensure the performance of wireless communication systems. Massive Multiple Input Multiple Output (M-MIMO) systems have a dramatic increase...
详细信息
This paper introduces a preconditioned method designed to comprehensively address the saddle point system with the aim of improving convergence efficiency. In the preprocessor construction phase, a technical approach ...
详细信息
In the Software Defined network (SDN), the terminal has been connected to the network in the process when an access point failure or damage, if the access point is not properly selected, it will lead to a decline in t...
详细信息
This technical note presents a counterexample showing that the equivalence conditions proposed by Geng et al. (IEEE Trans. Automat. Control, 2024), which use a minimum-order compensator (MOC) to achieve desired design...
详细信息
暂无评论