Federated learning (FL) has been demonstrated to be susceptible to backdoor attacks. However, existing academic studies on FL backdoor attacks rely on a high proportion of real clients with main task-related data, whi...
详细信息
We address the problem of learning new classes for semantic segmentation models from few examples, which is challenging because of the following two reasons. Firstly, it is difficult to learn from limited novel data t...
We address the problem of learning new classes for semantic segmentation models from few examples, which is challenging because of the following two reasons. Firstly, it is difficult to learn from limited novel data to capture the underlying class distribution. Secondly, it is challenging to retain knowledge for existing classes and to avoid catastrophic forgetting. For learning from limited data, we propose a pseudo-labeling strategy to augment the few-shot training annotations in order to learn novel classes more effectively. Given only one or a few images labeled with the novel classes and a much larger set of unlabeled images, we transfer the knowledge from labeled images to unlabeled images with a coarse-to-fine pseudo-labeling approach in two steps. Specifically, we first match each labeled image to its nearest neighbors in the unlabeled image set at the scene level, in order to obtain images with a similar scene layout. This is followed by obtaining pseudo-labels within this neighborhood by applying classifiers learned on the few-shot annotations. In addition, we use knowledge distillation on both labeled and unlabeled data to retain knowledge on existing classes. We integrate the above steps into a single convolutional neural network with a unified learning objective. Extensive experiments on the Cityscapes and KITTI datasets validate the efficacy of the proposed approach in the self-driving domain. Code is available from https://***/ChasonJiang/FSCILSS.
The emergence of Segment Routing(SR)provides a novel routing paradigm that uses a routing technique called source packet *** SR architecture,the paths that the packets choose to route on are indicated at the ingress *...
详细信息
The emergence of Segment Routing(SR)provides a novel routing paradigm that uses a routing technique called source packet *** SR architecture,the paths that the packets choose to route on are indicated at the ingress *** with shortest-path-based routing in traditional distributed routing protocols,SR can realize a flexible routing by implementing an arbitrary flow splitting at the ingress *** the advantages of SR,it may be difficult to update the existing IP network to a full SR deployed network,for economical and technical *** partial of the traditional IP network to the SR network,thus forming a hybrid SR network,is a preferable *** the traffic is dynamically changing in a daily time,in this paper,we propose a Weight Adjustment algorithm WASAR to optimize routing in a dynamic hybrid SR *** algorithm can be divided into three steps:firstly,representative Traffic Matrices(TMs)and the expected TM are obtained from the historical TMs through ultrascalable spectral clustering ***,given the network topology,the initial network weight setting and the expected TM,we can realize the link weight optimization and SR node deployment optimization through a Deep Reinforcement Learning(DRL)***,we optimize the flow splitting ratios of SR nodes in a centralized online manner under dynamic traffic demands,in order to improve the network *** the evaluation,we exploit historical TMs to test the performance of the obtained routing configuration in *** extensive experimental results validate that our proposed WASAR algorithm has superior performance in reducing Maximum Link Utilization(MLU)under the dynamic traffic.
Pulsar search is always the basis of pulsar navigation, gravitational wave detection and other research topics. Currently, the volume of pulsar candidates collected by Five-hundred-meter Aperture Spherical radio Teles...
详细信息
In intelligent transportation systems (ITSs), incorporating pedestrians and vehicles in-the-loop is crucial for developing realistic and safe traffic management solutions. However, there is falls short of simulating c...
详细信息
3D anomaly detection has recently become a significant focus in computer vision. Several advanced methods have achieved satisfying anomaly detection performance. However, they typically concentrate on the external str...
详细信息
Face forgery detection (FFD) is devoted to detecting the authenticity of face images. Although current CNN-based works achieve outstanding performance in FFD, they are susceptible to capturing local forgery patterns g...
详细信息
In an effort to the problem of insufficient tracking performance of the Fully-convolutional Siamese network (SiamFC) in complex scenarios, a dual attention mechanism object tracking algorithm based on the Fully-convol...
详细信息
Due to the domain shifts between training and testing medical images, learned segmentation models often experience significant performance degradation during deployment. In this paper, we first decompose an image into...
详细信息
In multi-site brain disease diagnosis studies, traditional centralized training methods necessitate sharing medical data, posing significant privacy risks. Federated learning (FL) offers a privacy-preserving solution ...
详细信息
ISBN:
(数字)9798350380323
ISBN:
(纸本)9798350380330
In multi-site brain disease diagnosis studies, traditional centralized training methods necessitate sharing medical data, posing significant privacy risks. Federated learning (FL) offers a privacy-preserving solution by enabling global model training through aggregating locally trained models from multiple data centers without sharing raw data. However, current FL approaches rely on a server-based network topology, where central server failure disrupts training. Additionally, data heterogeneity across sites often slows convergence and reduces accuracy. To overcome these issues, we introduce a decentralized personalized federated learning collaborative aggregation network (pFedCAN). This framework has two core components: (1) separating local models into shared and personalized layers, and (2) forming a collaborative aggregation network via similarity detection in the shared layers. Specifically, each center trains its local model, then separates it into shared and personalized layers. The shared layer is exchanged with other centers, while the personalized layer remains local. data centers analyze similarities in received shared layers to build a collaborative network, where shared layers from similar centers are aggregated to refine the model. This approach flexibly adapts to varying levels of data heterogeneity, enhancing model training efficiency. Validation on public datasets, ABIDE I and ADHD, shows that the proposed method outperforms current leading techniques.
暂无评论