Prototype-based federated learning has emerged as a promising approach that shares lightweight prototypes to transfer knowledge among clients with data heterogeneity in a model-agnostic manner. However, existing methods often collect prototypes directly from local models, which inevitably introduce inconsistencies into representation learning due to the biased data distributions and differing model architectures among clients. In this paper, we identify that both statistical and model heterogeneity create a vicious cycle of representation inconsistency, classifier divergence, and skewed prototype alignment, which negatively impacts the performance of clients. To break the vicious cycle, we propose a novel framework named Federated Learning via Semantic Anchors (FedSA) to decouple the generation of prototypes from local representation learning. We introduce a novel perspective that uses simple yet effective semantic anchors serving as prototypes to guide local models in learning consistent representations. By incorporating semantic anchors, we further propose anchor-based regularization with margin-enhanced contrastive learning and anchor-based classifier calibration to correct feature extractors and calibrate classifiers across clients, achieving intra-class compactness and inter-class separability of prototypes while ensuring consistent decision boundaries. We then update the semantic anchors with these consistent and discriminative prototypes, which iteratively encourage clients to collaboratively learn a unified data representation with robust generalization. Extensive experiments under both statistical and model heterogeneity settings show that FedSA significantly outperforms existing prototype-based FL methods on various classification tasks.

Low-light images are typically characterized by low pixel intensity and color distortion, presenting a significant challenge for accurate 3D reconstruction with 3D Gaussian Splatting (3DGS). Traditional 2D enhancement methods fail to maintain consistent illumination, affecting reconstruction quality. We propose Low-Light Gaussian (LLGS), which can directly leverage low-light images for 3D reconstruction and synthesizing normal-light novel views. LLGS incorporates absorptance to simulate light behavior in low-light conditions, assuming objects maintain normal illumination while reflected light intensity attenuates due to absorptance during rendering. This approach enables the capture of authentic color information in dimly lit scenes. LLGS outperforms current enhancement algorithms and Neural Radiance Fields (NeRF) in image quality and processing efficiency, making it highly effective for low-light 3D scene reconstruction and novel view synthesis.

End-edge collaborative inference refers to the fact that edge servers (ESs) and end devices (EDs) jointly participate in inference tasks, which can not only reduce communication latency and bandwidth consumption with the cloud but also protect user data privacy. However, existing collaborative inference methods do not fully consider the limited resources of EDs and ignore the latency and accuracy requirements of different inference tasks. In this paper, we design a DNN inference acceleration framework to balance inference latency and accuracy. Specifically, we first use a compression method based on deep reinforcement learning to determine the compression ratio and deeply compress the original model to reduce the complexity of the model. To reduce the cumulative error caused by compression, a knowledge distillation-based scheme is used to fine-tune the compressed model. Finally, the DNN model is partitioned and deployed on the ED and ES, respectively. Extensive experiments demonstrate the effectiveness of the framework in achieving lowlatency DNN inference on demand.

Dense text retrieval has become a widely used paradigm for recalling existing language models , and efficient dense text retrieval is essential for obtaining sufficiently accurate candidate samples. However, the existing methods for dense text retrieval, which typically use dual-encoder architectures to match similar queries and documents, suffer from a lack of information interaction at low data volumes, resulting in suboptimal performance. Additionally, existing research relies on negative sampling techniques that do not take into account the negative effects of single negative sampling bias on the robustness of the model. These limitations hinder the development of more robust dense text retrieval models . In this paper, we propose a multi-view contrast learning architecture, named MvCR, to address these issues. MvCR improves the performance of dense text retrieval by performing contrast learning with multiple views while significantly increasing the model’s ability to discriminate between positive and negative samples. Additionally, we propose a data augmentation method that focuses on increasing the number of hard negative samples with accurate and semantic matching features. The experimental results have shown that MvCR can perform as well as strong baseline models even when the data volume is small. Furthermore, MvCR achieved better results on two popular retrieval benchmarks with comparable amounts of data. Specifically, MRR@10 was 39 . 1 ( + 0 . 9 % ) and Recall@50 was 87 . 8 ( + 1 . 4 % ) on the MS-MARCO dataset. And Recall@5 increased to 77 . 2 ( + 1 . 8 % ) and Recall@50 increased to 85 . 3 ( + 1 . 0 % ) on the Natural Questions dataset.

The most significant advantage of the unseen relation extraction is that it can recognize unlabeled relations.While Zero-Shot Learning can meet the requirements of the identification of unseen relation through relation description information without labeled datasets.However, unseen relation extraction requires an effective method in representation and generalization, which become a challenge for zero-shot learning approach.In this paper, we propose a Zero-Shot learning Relation Extraction based on Contrastive learning Model (ZRCM) to capture deep interrelation text information.We design a comparison sample generation method which can produce several instances for one input sentence and compare the distance between positive instance and negative ones, so as to improve the hidden text information mining ability.Experiments conducted on relation extraction common datasets confirmed the promotion of ZRCM compared with the existing methods.Especially, our model can improve the F1 value by up to 7% at best.When there are fewer unseen relations to predict, our model can achieve better performance.

Different from existing cross-project defection prediction(CPDP) problems which assume that there is a close relation between the source data sets and the target data sets, in the heterogenous cross-project defection prediction(HCPDP) problem, the target data sets can be totally different from the source data sets. In order to narrow the difference between source data sets and target data sets, we implemented our own algorithm SLA + based on the selective learning algorithm . We select one of the multiple sources that have the highest similarity to the target data set as the source data set, and select one or more of the other source data sets that are similar to both the target data set and the source data set as an intermediate domain. We set up a bridge between the target domain and the source domain through the intermediate domain , breaking the large distribution gap for transferring knowledge between the source domain and the target domain. Besides, we achieve the purpose of dimensionality reduction by mining the potential relationship between features. We have done experiments on open source data sets, and the data sets used are all heterogeneous. The experiments prove that our method achieves comparable results compared with state-of-the-art HCPDP in most cases.