Long-tailed semi-supervised learning poses a significant challenge in training models with limited labeled data exhibiting a long-tailed label distribution. Current state-of-the-art LTSSL approaches heavily rely on high-quality pseudo-labels for large-scale unlabeled data. However, these methods often neglect the impact of representations learned by the neural network and struggle with real-world unlabeled data, which typically follows a different distribution than labeled data. This paper introduces a novel probabilistic framework that unifies various recent proposals in long-tail learning. Our framework derives the class-balanced contrastive loss through Gaussian kernel density estimation. We introduce a continuous contrastive learning method, CCL, extending our framework to unlabeled data using reliable and smoothed pseudo-labels. By progressively estimating the underlying label distribution and optimizing its alignment with model predictions, we tackle the diverse distribution of unlabeled data in real-world scenarios. Extensive experiments across multiple datasets with varying unlabeled data distributions demonstrate that CCL consistently outperforms prior state-of-the-art methods, achieving over 4% improvement on the ImageNet-127 dataset. Our source code is available at https://github.com/zhouzihao11/CCL
Underwater target classification is an important research topic in the field of underwater acoustic signal processing, and the classification of marine mammal vocalizations is of great significance for the conservation of them. In this paper, the calls of humpback whales, short-finned pilot whales and sperm whales are taken as the research objects and plotted as spectrograms. Aiming to address the issue of small targets in spectrograms, we proposed an improved model based on YOLOv5 and integrate the Vision Transformer BiFormer module to enhance the detection accuracy of each sample. The test result shows that the mean average precision of all categories in the improved model is 88.1%, a 4.9% increase compared to the pre-improvement performance.
Label-free microscopy that does not require the staining of weakly absorbing samples circumvent the adverse effects of exogenous dyes on the biological sample, and thus have received much attention in recent years. Among them, non-interferometric optical diffraction tomography microscopy has become a hot topic in the direction of label-free three-dimensional microscopy due to its system simplicity, ease of integration and independence from scattering noise. We recently designed a refractive index and fluorescence dual-modality microscopy system based on the transport of intensity diffraction tomography microscopy, which solves the absorption and phase information of the sample separately with the help of two illuminating apertures, and optimizes the imaging speed by applying an electrically tunable lens.
Temporal graph clustering is a complex task that involves discovering meaningful structures in dynamic graphs where relationships and entities change over time. Existing methods typically require centralized data collection, which poses significant privacy and communication challenges. In this work, we introduce a novel Federated Temporal Graph Clustering (FTGC) framework that enables decentralized training of graph neural networks (GNNs) across multiple clients, ensuring data privacy throughout the process. Our approach incorporates a temporal aggregation mechanism to effectively capture the evolution of graph structures over time and a federated optimization strategy to collaboratively learn high-quality clustering representations. By preserving data privacy and reducing communication overhead, our framework achieves competitive performance on temporal graph datasets, making it a promising solution for privacy-sensitive, real-world applications involving dynamic data.
Abstract CD8 + T cells are integral to the effective management of cancer and infectious diseases due to their cytotoxic functions. The efficacy of these cells is profoundly influenced by their metabolic state, which regulates their activation, differentiation, and longevity. Accordingly, the modulation of metabolic pathways within CD8 + T cells is crucial for enhancing the effectiveness of T cell-based immunotherapy. Precise metabolic control is paramount in optimizing therapeutic outcomes and minimizing potential toxicities associated with treatment. Importantly, the potential of exogenous metabolites to augment CD8 + T cell responses is critically evaluated, especially through in vivo evidence that underscores their therapeutic promise. This review also addresses current challenges, including the need for precise control of metabolic modulation to avoid adverse effects, the development of targeted delivery systems to ensure efficient metabolite delivery to CD8 + T cells, and the inherent variability of metabolic states among patients that may influence treatment outcomes. Addressing these hurdles will be crucial for the successful integration of metabolic interventions into established immunotherapeutic regimens.
Optical diffraction tomography (ODT) is an important technique for three-dimensional (3D) imaging of semi-transparent biological samples, enabling volumetric visualization of living cells, cultures, and tissues without the need for exogenous dyes. However, ODT faces significant challenges in imaging complex biological specimens due to the limited specificity of refractive index (RI) and the coupled relationship between absorption and phase in image formation. Here, we present multi-modal transport of intensity diffraction tomography (MM-TIDT), a high-speed 3D microscopy technique that integrates an electrically tunable lens with modified illumination patterns to decouple phase and absorption information. Leveraging dual illumination schemes-circular and annular apertures-MM-TIDT acquires two intensity stacks, facilitating accurate phase and absorption decoupling. Based on an alternating direction method of multipliers (ADMM) framework with total variation (TV) and non-negativity regularization, our method reconstructs multi-modal 3D distributions of fluorescence and complex RI with high accuracy and robustness. Experimental validation with fluorescent polystyrene microspheres, Spirulina specimens, and DAPI-labeled C166 cells demonstrates the multi-modal imaging capability of MM-TIDT to resolve fine structural details across diverse sample types, providing a versatile platform for exploring dynamic biological processes and intricate cellular interactions.
T cell receptor-engineered T (TCR-T) cell therapies are at the forefront of cancer immunotherapy, offering a transformative approach that significantly enhances the ability of T cells to recognize and eliminate cancer cells. This innovative method involves genetically modifying TCRs to increase their affinity for tumor-specific antigens. While these enhancements improve the ability of T cells to recognize and bind to antigens on cancer cells, rigorous assessment of specificity remains crucial to ensure safety and targeted responses. This dual focus on affinity and specificity holds significant promise for the treatment of solid tumors, enabling precise and efficient cancer cell recognition. Despite rapid advancements in TCR engineering and notable progress in TCR screening technologies, as evidenced by the growing number of specific TCRs entering clinical trials, several technical and clinical challenges remain. These challenges primarily pertain to the specificity, affinity, and safety of engineered TCRs. Moreover, the accurate identification and selection of TCRs that are both effective and safe are essential for the success of TCR-T cell therapies in cancer treatment. This review provides a comprehensive examination of the theoretical foundations of TCR therapy, explores strategies for screening specific TCRs and antigens, and highlights the ongoing challenges in this evolving therapeutic landscape.
Personalized diagnosis prediction based on electronic health records (EHR) of patients is a promising yet challenging task for AI in healthcare. Existing studies typically ignore the heterogeneity of diseases across different patients. For example, diabetes can have different complications across different patients (e.g., hyperlipidemia and circulatory disorder), which requires personalized diagnoses and treatments. Specifically, existing models fail to consider 1) varying severity of the same diseases for different patients, 2) complex interactions among syndromic diseases, and 3) dynamic progression of chronic diseases. In this work, we propose to perform personalized diagnosis prediction based on EHR data via capturing disease severity, interaction, and progression. In particular, we enable personalized disease representations via severity-driven embeddings at the disease level. Then, at the visit level, we propose to capture higher-order interactions among diseases that can collectively affect patients' health status via hypergraph-based aggregation; at the patient level, we devise a personalized generative model based on neural ordinary differential equations to capture the continuous-time disease progressions underlying discrete and incomplete visits. Extensive experiments on two real-world EHR datasets show significant performance gains brought by our approach, yielding average improvements of 10.70% for diagnosis prediction over state-of-the-art competitors.
Graphs are widely used to model interconnected entities and improve downstream predictions in various real-world applications. However, real-world graphs nowadays are often associated with complex attributes on multiple types of nodes and even links that are hard to model uniformly, while the widely used graph neural networks (GNNs) often require sufficient training toward specific downstream predictions to achieve strong performance. In this work, we take a fundamentally different approach than GNNs, to simultaneously achieve deep joint modeling of complex attributes and flexible structures of real-world graphs and obtain unsupervised generic graph representations that are not limited to specific downstream predictions. Our framework, built on a natural integration of language models (LMs) and random walks (RWs), is straightforward, powerful and data-efficient. Specifically, we first perform attributed RWs on the graph and design an automated program to compose roughly meaningful textual sequences directly from the attributed RWs; then we fine-tune an LM using the RW-based textual sequences and extract embedding vectors from the LM, which encapsulates both attribute semantics and graph structures. In our experiments, we evaluate the learned node embeddings towards different downstream prediction tasks on multiple real-world attributed graph datasets and observe significant improvements over a comprehensive set of state-of-the-art unsupervised node embedding methods. We believe this work opens a door for more sophisticated technical designs and empirical evaluations toward the leverage of LMs for the modeling of real-world graphs.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)