Photonic crystal fiber (PCF) inverse design remains challenging because candidate geometries must satisfy coupled optical targets under expensive electromagnetic simulation. Existing pipelines improve surrogate prediction or one-shot parameter recommendation, but they do not accumulate reusable design knowledge across iterative trials. We formulate PCF inverse design as a memory-policy learning problem and propose SkillPCF, a closed-loop agent framework that combines a physics-guided memory skill bank, reinforcement-learned skill selection, and simulator-grounded skill evolution. We further construct a real-world dataset with 479 expert interaction traces (2507 spans) and 553 memory-dependent evaluation queries covering dispersion engineering, loss optimization, and multi-objective design. Experiments across multiple LLM backbones and classical baselines show that SkillPCF achieves stronger design-quality and efficiency trade-offs under practical simulation budgets, demonstrating the effectiveness of our proposed memory-skill learning paradigm for physics-aware PCF inverse design.

Training a general-purpose time series foundation models with robust generalization capabilities across diverse applications from scratch is still an open challenge. Efforts are primarily focused on fusing cross-domain time series datasets to extract shared subsequences as tokens for training models on Transformer architecture. However, due to significant statistical heterogeneity across domains, this cross-domain fusing approach doesn't work effectively as the same as fusing texts and images. To tackle this challenge, this paper proposes a novel federated learning approach to address the heterogeneity in time series foundation models training, namely FFTS. Specifically, each data-holding organization is treated as an independent client in a collaborative learning framework with federated settings, and then many client-specific local models will be trained to preserve the unique characteristics per dataset. Moreover, a new regularization mechanism will be applied to both client-side and server-side, thus to align the shared knowledge across heterogeneous datasets from different domains. Extensive experiments on benchmark datasets demonstrate the effectiveness of the proposed federated learning approach. The newly learned time series foundation models achieve superior generalization capabilities on cross-domain time series analysis tasks, including forecasting, imputation, and anomaly detection.

Training federated foundation models (FFMs) for sensitive medical images presents open challenges due to complex data heterogeneity. Related studies focus on the problem within a single imaging modality; however, these methods lacks the flexibility required to generalize related tasks uniformly across different imaging modalities. This paper proposes FFMed, a federated learning framework that tames pretrained Vision-Language Models for FFMs training, targeting medical image classification across heterogeneous medical imaging modalities. Specifically, FFMed improves the CLIP's medical image-text alignment through Adaptive Prompt Generation to introduce task/domain-specific, informative prompts in conjunction with low-rank adaption. To mitigate learning bias from imaging modalities heterogeneity across clients, we propose Anchor-based Dynamic Regularization, which dynamically constrains local optimization to remain close to the global stationary point, thereby promoting optimal global consensus. Ultimately, FFMed fosters a unified model that effectively generalizes across diverse non-IID environments. Extensive experiment on real-world medical image datasets demonstrate the effectiveness and superiority of FFMed.

Remaining useful life (RUL) prediction of bearings is essential for effective prognostics and health management (PHM). Although deep learning-based RUL prediction methods achieve high prediction accuracy, they often introduce significant parameter redundancy due to their inability to efficiently capture the intricate temporal dynamics in bearing degradation signals, leading to computationally expensive models with limited practical applicability. To address this challenge, we propose a novel RUL prediction framework that integrates the Wasserstein distance of cyclic spectrum (WDCS) with a Lightweight TimesNet (WDCS-LTN). Specifically, the WDCS serves as a health indicator, effectively extracting multiperiodic features from bearing degradation signals. Subsequently, the LTN transforms the 1-D WDCS sequence into multiple 2-D tensors with varying localities, enabling precise modeling of intraperiod and interperiod temporal dynamics. A shared lightweight inception block is constructed within the LTN to capture temporal variations in 2-D space while maintaining low model complexity. Experimental results on bearing degradation datasets show that WDCS-LTN achieves a prediction error (mean absolute error) of 0.091 with only 37k parameters, outperforming existing methods in terms of accuracy, parameter efficiency, and memory consumption. Through efficiently modeling the temporal dynamics, WDCS-LTN ensures practicality for industrial applications by addressing parameter redundancy while offering enhanced prediction capabilities.

This paper demonstrates that pre-trained language models (PLMs) are strong foundation models for on-device meteorological variable modeling. We present LM-Weather, a generic approach to taming PLMs, that have learned massive sequential knowledge from the universe of natural language databases, to acquire an immediate capability to obtain highly customized models for heterogeneous meteorological data on devices while keeping high efficiency. Concretely, we introduce a lightweight personalized adapter into PLMs and endows it with weather pattern awareness. During communication between clients and the server, low-rank-based transmission is performed to effectively fuse the global knowledge among devices while maintaining high communication efficiency and ensuring privacy. Experiments on real-wold dataset show that LM-Weather outperforms the state-of-the-art results by a large margin across various tasks (e.g., forecasting and imputation at different scales). We provide extensive and in-depth analyses experiments, which verify that LM-Weather can (1) indeed leverage sequential knowledge from natural language to accurately handle meteorological sequence, (2) allows each devices obtain highly customized models under significant heterogeneity, and (3) generalize under data-limited and out-of-distribution (OOD) scenarios.

Remote Sensing Target Fine-grained Classification (TFGC) is of great significance in both military and civilian fields. Due to location differences, growth in data size, and centralized server storage constraints, these data are usually stored under different databases across regions/countries. However, privacy laws and national security concerns constrain researchers from accessing these sensitive remote sensing images for further analysis. Moreover, low-resource remote sensing devices face challenges in communication overhead and efficiency when dealing with the ever-increasing data and model scales. To address these challenges, this paper proposes a novel P rivacy- R eserving TFGC F ramework based on Federated L earning, dubbed PRFL . The proposed framework allows each client to learn global and local knowledge to enhance the local representation of private data in environments with extreme statistical heterogeneity ( non. Independent and Identically Distributed, IID ). It provides highly customized models to clients with differentiated data distributions. Furthermore, the framework minimizes communication overhead and improves efficiency while ensuring satisfactory performance, enhancing robustness and practical applicability under resource-scarce conditions. We demonstrate the effectiveness of PRFL on the classical TFGC task using four public datasets.

Chest imaging plays an essential role in diagnosing and predicting patients with COVID-19 with evidence of worsening respiratory status. Many deep learning-based approaches for pneumonia recognition have been developed to enable computer-aided diagnosis. However, the long training and inference time makes them inflexible, and the lack of interpretability reduces their credibility in clinical medical practice. This paper aims to develop a pneumonia recognition framework with interpretability, which can understand the complex relationship between lung features and related diseases in chest X-ray (CXR) images to provide high-speed analytics support for medical practice. To reduce the computational complexity to accelerate the recognition process, a novel multi-level self-attention mechanism within Transformer has been proposed to accelerate convergence and emphasize the task-related feature regions. Moreover, a practical CXR image data augmentation has been adopted to address the scarcity of medical image data problems to boost the model's performance. The effectiveness of the proposed method has been demonstrated on the classic COVID-19 recognition task using the widespread pneumonia CXR image dataset. In addition, abundant ablation experiments validate the effectiveness and necessity of all of the components of the proposed method.

To tackle the global climate challenge, it urgently needs to develop a collaborative platform for comprehensive weather forecasting on large-scale meteorological data. Despite urgency, heterogeneous meteorological sensors across countries and regions, inevitably causing multivariate heterogeneity and data exposure, become the main barrier. This paper develops a foundation model across regions capable of understanding complex meteorological data and providing weather forecasting. To relieve the data exposure concern across regions, a novel federated learning approach has been proposed to collaboratively learn a brand-new spatio-temporal Transformer-based foundation model across participants with heterogeneous meteorological data. Moreover, a novel prompt learning mechanism has been adopted to satisfy low-resourced sensors' communication and computational constraints. The effectiveness of the proposed method has been demonstrated on classical weather forecasting tasks using three meteorological datasets with multivariate time series.

Meteorological radar reflectivity data (i.e. radar echo) significantly influences precipitation prediction. It can facilitate accurate and expeditious forecasting of short-term heavy rainfall bypassing the need for complex Numerical Weather Prediction (NWP) models. In comparison to conventional models, Deep Learning (DL)-based radar echo extrapolation algorithms exhibit higher effectiveness and efficiency. Nevertheless, the development of reliable and generalized echo extrapolation algorithm is impeded by three primary challenges: cumulative error spreading, imprecise representation of sparsely distributed echoes, and inaccurate description of non-stationary motion processes. To tackle these challenges, this paper proposes a novel radar echo extrapolation algorithm called Temporal-Spatial Parallel Transformer, referred to asTempEE.TempEEavoids using auto-regression and instead employs a one-step forward strategy to prevent cumulative error spreading during the extrapolation process. Additionally, we propose the incorporation of a Multi-level Temporal-Spatial Attention mechanism to improve the algorithm’s capability of capturing both global and local information while emphasizing task-related regions, including sparse echo representations, in an efficient manner. Furthermore, the algorithm extracts spatio-temporal representations from continuous echo images using a parallel encoder to model the non-stationary motion process for echo extrapolation. The superiority of ourTempEEhas been demonstrated in the context of the classic radar echo extrapolation task, utilizing a real-world dataset. Extensive experiments have further validated the efficacy and indispensability of various components withinTempEE.

Typhoons, a kind of devastating natural disaster, have caused incalculable damages worldwide. The meteorological radar image is essential for weather forecasting, especially typhoons. The weather nowcasting (future 0-6 hours) can be implemented via extrapolating radar images without using the primary weather forecasting method – the Numerical Weather Prediction model. However, the existing related techniques based on statistics or Artificial Intelligence were not efficient enough. In this paper, a novel radar image extrapolation algorithm named Dynamic Multi-Scale Fusion-Generative Adversarial Network (DMSF-GAN) was proposed. DMSF-GAN captures the future radar image distribution based on current radar images through modifying the GAN. In the generative module of the GAN, an auto-encoder consisting of Dynamic Inception-3D and Feature Connection blocks extracts significant features from current radar images. The feasibility of the proposed model was verified on a real radar image dataset, and experimental results proved that the proposed algorithm could effectively capture the location and pattern of the future radar echo, especially for typhoon weather systems. Compared with mainstream methods of radar image extrapolation such as Optical-Flow and Recurrent Neural Network (RNN)-based models, DMSF-GAN has a more superior and robust performance, which is also suitable for running on low configuration machines.