Semantic communication is an emerging paradigm for digital twin (DT) synchronization in unmanned aerial vehicle (UAV)-assisted edge computing environments, where machine learning (ML) models are deployed on edge servers and UAVs as semantic encoders and decoders to perform real-time synchronization. However, with limited system resources, additional computation workloads are still brought to all participants for semantic information extraction and recovery. In this work, we propose an optimized tiny ML-based DT synchronization framework to minimize the synchronization latency in UAV-assisted edge computing environments, considering time-average constraints on virtual energy deficit queue stability. Due to the coexistence of tiny ML-based semantic communications, a semantic extraction factor is introduced to formulate the DT synchronization problem as a time-average time minimization problem. By leveraging the Lyapunov optimization framework, the multi-stage DT synchronization problem is transformed into several per-slot resource allocation problems. To solve the per-slot optimization problem efficiently, a deep reinforcement learning-based synchronization (DRLS) algorithm is proposed, where an actor-critic structure is adopted to generate synchronization actions with low time complexity. Finally, we conduct simulation experiments to evaluate the performance of the proposed DRLS scheme. Numerical results demonstrate that our DRLS algorithm can reduce 8.23% of DT synchronization delay and 15.31% of synchronization data dropping rates on average by comparing it with the UAV-edge collaborative synchronization scheme without semantic communications. Besides, the DRLS algorithm can achieve up to 57.14% synchronization energy reduction compared with representative synchronization policies.

Long-tailed (LT) classification is an unavoidable and challenging problem in the real world. Most existing long-tailed classification methods focus only on solving the class-wise imbalance while ignoring the attribute-wise imbalance. The deviation of a classification model is caused by both class-wise and attribute-wise imbalance. Due to the fact that attributes are implicit in most datasets and the combination of attributes is complex, attribute-wise imbalance is more difficult to handle. For this purpose, we proposed a novel long-tailed classification framework, aiming to build a multi-granularity classification model by means of invariant feature learning. This method first unsupervisedly constructs Coarse-Grained Leading Forest (CLF) to better characterize the distribution of attributes within a class. Depending on the distribution of attributes, one can customize suitable sampling strategies to construct different imbalanced datasets. We then introduce multi-center loss (MCL) that aims to gradually eliminate confusing attributes during feature learning process. The proposed framework does not necessarily couple to a specific LT classification model structure and can be integrated with any existing LT method as an independent component. Extensive experiments show that our approach achieves state-of-the-art performance on both existing benchmarks ImageNet-GLT and MSCOCO-GLT and can improve the performance of existing LT methods.

The distribution of the labeled data can greatly affect the performance of a semi-supervised learning (SSL) model. Most existing SSL models select the labeled data randomly and equally allocate the labeling quota among the classes, leading to considerable unstableness and degeneration of performance. This study unsupervisedly constructs a leading forest that forms another metric space, based on which it is convenient to define the fuzzy membership function to characterize central and divergent samples and select both types with fuzzy Xor logic. The labeling quota can thus be allocated adaptively among different classes. The proposed determinate labeling strategy can generally improve the performance for most SSLs. Especially, when combined with the kernelized large margin component analysis, it produces a novel semi-supervised classification model. In addition, the multi-modal issue in SSL is effectively addressed by the multi-granular structure of leading forest that readily facilitates multiple local metrics learning. Extensive experimental results demonstrate that the proposed method achieved competitive efficiency and encouraging accuracy when compared with the state-of-the-art methods

Graph contrastive learning (GCL) has emerged as a dominant paradigm for self‐supervised representation learning for attributed graph data. However, existing GCL methods heavily rely on empirical graph data augmentation, which may distort intrinsic graph semantics and produce poor generalisation without carefully chosen or designed augmentation techniques. Furthermore, most GCL approaches focus on same‐granularity contrastive learning (e.g., node vs. node), neglecting the hierarchical and multigranular properties inherent in real‐world networks, leading to suboptimal performance. To address these limitations, we propose HPoolGCL, a cross‐granularity GCL framework compatible with various hierarchical graph pooling methods to capture multigranularity information. Our framework eliminates the need for handcrafted augmentations, explicit negative sampling and complex multiencoder architectures by applying two novel loss functions in hierarchical graph pooling. The theoretical analysis is provided to explain the effectiveness of unified MGC and HiCR losses from three perspectives, namely, the information maximisation principle, the redundancy reduction principle and the information bottleneck principle. The experimental results demonstrate that HPoolGCL achieves state‐of‐the‐art performance across multiple downstream tasks on five benchmarks. Our codes are available at https://github.com/Heycen/HPoolGCL .