We present Flow-Induced Diagonal Gaussian Processes (FiD-GP), a compression framework that incorporates a compact inducing weight matrix to project a neural network’s weight uncertainty into a lower-dimensional subspace. Critically, FiD-GP relies on normalising flow variational posterior and spectral regularisations to augment its expressiveness and align the inducing subspace with feature-gradient geometry through a numerically stable projection mechanism objective. Furthermore, we demonstrate how the prediction framework in FiD-GP can help to design a single pass projection for Out-of-Distribution (OoD) detection. Our analysis shows that FiD-GP improves uncertainty estimation ability on various tasks compared with SVGP-based baselines, satisfies tight spectral residual bounds with theoretically guaranteed OoD detection, and significantly compresses the neural network’s storage requirements at the cost of increased inference computation dependent on the number of inducing weights employed. Specifically, in a comprehensive empirical study spanning regression, image classification, semantic segmentation, and Out-of-Distribution detection benchmarks, it significantly cuts Bayesian training cost, compresses parameters by roughly 51%, reduces model size by about 75%, and matches state-of-the-art accuracy and uncertainty estimation.

With the advent of the Industry 4.0 era, Federated Learning (FL) provides robust data privacy protection for smart manufacturing and supply chain optimization, while facilitating collaborative intelligent optimization across enterprises and devices. However, the complex and overparameterized deep neural networks used in FL result in significant computational overhead for Industrial Internet of Things (IIoT) devices, leading to low energy efficiency and hindering the practical deployment of FL on IIoT devices. Moreover, the widespread data and device heterogeneity in the IIoT exacerbates the decrease in energy efficiency caused by inconsistent computational efficiency across nodes. This article proposes an energy-efficient dynamic model pruning method for FL, named EDPrune-FL, to address the aforementioned challenges. Compared to existing methods, this approach offers greater flexibility and efficiency by utilizing a dynamic pruning rate allocation mechanism. This mechanism updates the pruning rate for each participating client in every communication round, allowing the pruning upper bound to adapt to the varying importance of different learning stages in FL. EDPrune-FL ensures the global model’s performance while reducing the training energy consumption of clients in heterogeneous environments. To guarantee that dynamic pruning maintains the stability and effectiveness of the model in heterogeneous environments, we also demonstrated the convergence of EDPrune-FL and discussed the relationship between pruning rates and convergence, providing a qualitative analysis. Experimental results demonstrate that our method outperforms the state-of-the-art technique across four real-world datasets. With tests conducted on 100 clients, our approach reduces energy consumption by 10% while maintaining comparable accuracy.

Shuhao Guan, Moule Lin, Cheng Xu, Xinyi Liu, Jinman Zhao, Jiexin Fan, Qi Xu, Derek Greene. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

While offering a principled framework for uncertainty quantification in deep learning, the employment of Bayesian Neural Networks (BNNs) is still constrained by their increased computational requirements and the convergence difficulties when training very deep, state-of-the-art architectures. In this work, We reinterpret weight-sharing quantization techniques from a stochastic perspective in the context of training and inference with Bayesian Neural Networks (BNNs). Specifically, we leverage 2D-adaptive Gaussian distributions, Wasserstein distance estimations, and alpha-blending to encode the stochastic behavior of a BNN in a lower-dimensional, soft Gaussian representation. Through extensive empirical investigation, we demonstrate that our approach significantly reduces the computational overhead inherent in Bayesian learning by several orders of magnitude, enabling efficient Bayesian training of large-scale models, such as ResNet-101 and Vision Transformer (VIT). On various computer vision benchmarks—including CIFAR-10, CIFAR-100, and ImageNet1k—our approach compresses model parameters by approximately 50$\times$ and reduces model size by 75% while achieving accuracy and uncertainty estimations comparable to state-of-the-art.

Federated learning (FL) has emerged as a pivotal collaborative machine learning framework, enabling privacy-preserving analytics for smart city applications using distributed data from Internet of Things (IoT) devices. However, the inherent data heterogeneity that arises from diverse geographical and environmental factors poses significant challenges to the effectiveness of FL-based models. To address these challenges, this paper introduces a novel Prototype-Based FL framework for cross-domain object extraction in heterogeneous remote sensing images. The proposed framework employs multiple vectors to represent class prototypes for capturing the intricate intra-class variations and mitigating the adverse effects of non-identically distributed (non-IID) data across clients. Furthermore, we adopt a distance-based classification method to reduce classification errors. Additionally, we propose a Prototype-Anchored Metric Learning approach to minimize intra-class variance and enhance inter-class separability, which can facilitate the alignment of feature representations across heterogeneous datasets. The proposed method improves the coherence and stability of feature spaces in federated settings and enhances the global model’s generalization capabilities for complex urban monitoring tasks. Extensive experiments on three distinct remote sensing datasets(including infrastructure and disaster) demonstrate that the proposed method significantly outperforms state-of-the-art FL-based approaches in urban monitoring accuracy and robustness. The code is available at

Accurate extraction of building vector boundaries holds paramount importance within the domains of urban planning and Geographic Information Systems (GIS), providing indispensable support for urban construction endeavors and resource management initiatives. CNNs, while proficient in local feature extraction, often falter in capturing holistic, global image characteristics. Transformers excel in contextual feature comprehension but demand substantial computational resources and parameterization, impeding practical deployment. To address these challenges, this paper introduces an innovative computational operator known as G-Swap, which integrates Gaussian-distance-based feature correlation considerations, thereby significantly augmenting contextual comprehension within the computational framework. Additionally, a universal architecture for boundary vector extraction is proposed in this paper, comprising three primary components: 1) an Enhanced Backbone, integrating the G-Swap operator to enhance the backbone while bolstering model expressiveness; 2) a Decoder module, tasked with discriminating corner and edge features; and 3) a Two-branch Detection Head. Empirical experiments conducted on the Vectorizing World Building Dataset (VWB) underscore the model’s superior performance. Our G-Swap achieved F1 scores of 91.2% for vertices and 80.1% for edges, surpassing the previous state-of-the-art by 2.1% and 2.0% respectively.

Effective forest tree species (TS) classification is critical for various application domains such as forest management, biodiversity conservation, and ecological research. However, existing studies on TS classification predominantly rely on high-cost and processing-intensive hyperspectral data, which limits practical applications on large scales. In this work, we focus on investigating the potential of cost-effective unmanned aerial vehicle (UAV) RGB images for TS classification in heterogeneous forests and propose a method that fully leverages the rich spatial, semantic, and visible spectral information of UAV RGB images. We propose an RSVMamba model, which incorporates improved visual state-space (VSS) blocks and an AutoDownsampling module to enhance accuracy and stability while paying particular attention to small objects in sparse spatial locations. The model achieves linear computational complexity while retaining the global receptive field, making it particularly suitable for processing high spatial-resolution images. Additionally, we collected UAV RGB images covering$40~\text {km}^{2}$of subtropical forest in southern China. A meticulous evaluation of this data shows that our method achieves an overall accuracy (OA) of 84.28% for eight TS, dead trees, and other broadleaves. We verify the superiority of our method through a series of comparative experiments on the collected and benchmark datasets. Our results affirm the usefulness of single-temporal UAV RGB images for TS classification in heterogeneous forest environments. Furthermore, the proposed method bridges the gap between data accessibility and precision in TS classification, broadening the boundaries of single-temporal UAV RGB images for practical forestry applications and providing a more cost-effective and time-flexible solution for this problem.

Post-OCR technology is used to correct errors in the text produced by OCR systems.This study introduces a method for constructing post-OCR synthetic data with different noise levels using weak supervision.We define Character Error Rate (CER) thresholds for "effective" and "ineffective" synthetic data, allowing us to create more useful multi-noise level synthetic datasets.Furthermore, we propose Self-Correct-Noise Test-Time Adaptation (SCN-TTA), which combines self-correction and noise generation mechanisms.SCN-TTA allows a model to dynamically adjust to test data without relying on labels, effectively handling proper nouns in long texts and further reducing CER.In our experiments we evaluate a range of models, including multiple PLMs and LLMs.Results indicate that our method yields models that are effective across diverse text types.Notably, the ByT5 model achieves a CER reduction of 68.67% without relying on manually annotated data 1 .

The building planar graph reconstruction, a.k.a. footprint reconstruction, which lies in the domain of computer vision and geoinformatics, has been long afflicted with the challenge of redundant parameters in conventional convolutional models. Therefore, in this letter, we proposed an advanced and adaptive shift architecture, the “Switch” operator, which incorporates nonexponential growth parameters while retaining analogous functionalities to integrate local feature spatial information, resembling a high-dimensional convolution operation. The “Switch” operator, cross-channel operation, architecture implements the XOR operation to exchange adjacent or diagonal features alternately and then blends alternating channels through a$1 \times 1$convolution operation to consolidate information from different channels. The SwitchNN architecture, on the other hand, incorporates a group-based parameter-sharing mechanism inspired by the convolutional neural network (CNN) process, thereby significantly reducing the number of parameters. We validated our proposed approach through experiments on the SpaceNet corpus. Our method achieves 82.9% precision, 79.8% F1 score, 83.7% recall, and an MAE of 0.018 with the FLOPs of 24.83 G, outperforming existing state-of-the-art methods, such as Roof-Former and HEAT. These results demonstrate the effectiveness of this innovative architecture in building planar graph reconstruction from 2-D building images.

Hyperspectral image (HSI) registers hundreds of spectral bands, whose intraclass variability and interclass similarity are resourceful information to be mined. Intraclass variability reflects the nonuniform and redundancy of the spatial and semantic features extracted from HSI. Interclass similarity represents the inherent relationship between adjacent features and snapshots. Existing models extract the superficial correlation representation for HSI to tackle the classification task but fail to embed the interclass and intraclass correlations due to these models’ intrinsic bottlenecks. Confronting the challenges of capturing interrelation for complex data in practice, we propose a Context-Aware Attentional Graph U-Net (CAGU) to improve these two modes of representation, which is more flexible in feature enhancement. In this method, attentional Graph U-Net is capable of extracting the intraclass embeddings within a non-Euclidean space by combining similar distributing feature vertices. The gated recurrent unit (GRU) is another critical component of our model to capture the context-aware dynamic interclass embeddings. Extensive experiments demonstrate that our model can efficiently outperform state-of-the-art methods across-the-board on five wide-adopted public data sets, namely, Pavia University, Indian Pines, Salinas Scene-show, Houston 2013, and Houston 2018, on par with the same scale of model parameters.

Hyperspectral images (HSIs) have significant applications in various domains, since they register numerous semantic and spatial information in the spectral band with spatial variability of spectral signatures. Two critical challenges in identifying pixels of the HSI are, respectively, representing the correlated information among the local and global, as well as the abundant parameters of the model. To tackle this challenge, we propose a multi-scale U-shape multi-layer perceptron (MUMLP) a model consisting of the designed multi-scale channel (MSC) block and the U-shape multi-layer perceptron (UMLP) structure. MSC transforms the channel dimension and mixes spectral band feature to embed the deep-level representation adequately. UMLP is designed by the encoder–decoder structure with multi-layer perceptron layers, which is capable of compressing the large-scale parameters. Extensive experiments are conducted to demonstrate that our model can outperform state-of-the-art methods across the board on three wide-adopted public datasets, namely Pavia University (PaviaU), Houston 2013, and Houston 2018.