Generative modeling of graphs with spatial structure is essential across many applications from computer graphics to spatial genomics. Recent flow-based generative models have achieved impressive results by gradually adding and then learning to remove noise from these graphs. Existing models, however, use graph neural network architectures that are independent of the noise level, limiting their expressiveness. To address this issue, we introduce *Noise-Conditioned Graph Networks* (NCGNs), a class of graph neural networks that dynamically modify their architecture according to the noise level during generation. Our theoretical and empirical analysis reveals that as noise increases, (1) graphs require information from increasingly distant neighbors and (2) graphs can be effectively represented at lower resolutions. Based on these insights, we develop Dynamic Message Passing (DMP), a specific instantiation of NCGNs that adapts both the range and resolution of message passing to the noise level. DMP consistently outperforms noise independent architectures on a variety of domains including $3$D point clouds, spatiotemporal transcriptomics, and images.

Foundation models (FMs) have shown great promise in single-cell genomics, yet current approaches, such as scGPT, Geneformer, and scFoundation, rely on centralized training and language modeling objectives that overlook the tabular nature of single-cell data and raise significant privacy concerns. We present TABULA, a foundation model designed for single-cell transcriptomics, which integrates a novel tabular modeling objective and federated learning framework to enable privacy-preserving pretraining across decentralized datasets. TABULA directly models the cell-by-gene expression matrix through column-wise gene reconstruction and row-wise cell contrastive learning, capturing both gene-level relationships and cell-level heterogeneity without imposing artificial gene sequence order. Extensive experiments demonstrate the effectiveness of TABULA: despite using only half the pretraining data, TABULA achieves state-of-the-art performance across key tasks, including gene imputation, perturbation prediction, cell type annotation, and multi-omics integration. It is important to note that as public single-cell datasets continue to grow, TABULA provides a scalable and privacy-aware foundation that not only validates the feasibility of federated tabular modeling but also establishes a generalizable framework for training future models under similar privacy-preserving settings.

Time-resolved single-cell omics data offers high-throughput, genome-wide measurements of cellular states, which are instrumental to reverse-engineer the processes underpinning cell fate. Such technologies are inherently destructive, allowing only cross-sectional measurements of the underlying stochastic dynamical system. Furthermore, cells may divide or die in addition to changing their molecular state. Collectively these present a major challenge to inferring realistic biophysical models. We present a novel approach, unbalanced probability flow inference, that addresses this challenge for biological processes modelled as stochastic dynamics with growth. By leveraging a Lagrangian formulation of the Fokker-Planck equation, our method accurately disentangles drift from intrinsic noise and growth. We showcase the applicability of our approach through evaluation on a range of simulated and real single-cell RNA-seq datasets. Comparing to several existing methods, we find our method achieves higher accuracy while enjoying a simple two-step training scheme.

For DC microgrids (MGs), real-time adjustment of current sharing ratios and secure voltage restoration are paramount for optimizing load allocation and enhancing dynamic performance. In this paper, a dual-objective distributed model-free adaptive control (MFAC) scheme is designed for the first time to guarantee voltage transient performance and adjustable current sharing. First, an output-constrained nonlinear MG model with ZIP (constant impedance, constant current and constant power) load is established, and subsequently it is converted into an equivalent unconstrained data model using system transformation and dynamic linearization techniques. Second, a new prescribed performance control algorithm with asymmetrical preset boundaries is proposed to restrict voltage transient responses. This algorithm is updated with real-time input and output data at discrete instants, making it independent of line resistance and ZIP load measurements. To enhance the robustness of the control method, an internal observer is designed to actively compensate for the unknown nonlinear dynamics generated by time-varying system parameters. The stability conditions of the transformed systems in the presence of ZIP loads and time-varying line resistance are derived, which can indirectly ensure the prescribed voltage performance of the original system. Finally, the effectiveness of the proposed control method is validated through some simulations and hardware experiments.

Semi-supervised learning for change detection (CD), which significantly reduces the labor costs associated with data annotation, has recently garnered substantial attention. In this study, we propose to enhance traditional semi-supervised learning frameworks by leveraging cross-task cross-domain (CTCD) models, which generate complementary features that differ from standard hidden features. The procedure is as follows. First, the standard features obtained from a traditional encoding–decoding structure are fused with attention-augmented complementary features. Second, a secondary decoder maps the fused heterogeneous features into the label space to obtain high-quality pseudo-labels, offering more precise guidance for semi-supervised learning on traditional structures. This approach improves pseudo-labels by leveraging the strength of CTCD models, including large pretrained models, to enhance the semi-supervised learning process of domain-specific and task-specific models. Experimental results on benchmark datasets demonstrate that our proposed approach surpasses state-of-the-art methods.

The time-resolved scRNA-seq (tscRNA-seq) provides the possibility to infer physically meaningful kinetic parameters, e.g., the transcription, splicing or RNA degradation rate constants with correct magnitudes, and RNA velocities by incorporating temporal information. Previous approaches utilizing the deterministic dynamics and steady-state assumption on gene expression states are insufficient to achieve favorable results for the data involving transient process. We present a dynamical approach, Storm (Stochastic models of RNA metabolic-labeling), to overcome these limitations by solving stochastic differential equations of gene expression dynamics. The derivation reveals that the new mRNA sequencing data obeys different types of cell-specific Poisson distributions when jointly considering both biological and cell-specific technical noise. Storm deals with measured counts data directly and extends the RNA velocity methodology based on metabolic labeling scRNA-seq data to transient stochastic systems. Furthermore, we relax the constant parameter assumption over genes/cells to obtain gene-cell-specific transcription/splicing rates and gene-specific degradation rates, thus revealing time-dependent and cell-state-specific transcriptional regulations. Storm will facilitate the study of the statistical properties of tscRNA-seq data, eventually advancing our understanding of the dynamic transcription regulation during development and disease.

Point clouds provide rich geometric representations, and point cloud semantic segmentation is essential in many applications. As the data scale of point clouds is usually quite large, some approaches propose constructing superpoint graphs from point clouds to reduce the time and space complexity during analysis. However, traditional superpoint-based graph neural network approaches for point cloud analysis typically aggregate features of adjacent superpoints and consider the most prominent feature only within each receptive field. In this work, we argue that adaptive varying-distance feature aggregation and discrimination can improve the effect of point cloud semantic segmentation. The proposed approach consists of three steps. First, we agglomerate points into superpoints and construct a superpoint graph as many traditional approaches. Second, we propose a novel varying-distance autoencoder to help each superpoint adaptively assimilate information from different distances. Third, we propose a discrimination loss to constrain the embedding space so that superpoints belonging to the same semantic class can get closer and vice versa. Regarding mIoU, our method outperforms the baseline by at least 8.1% for the S3DIS dataset and 3.1% in mIoU for the vKITTI dataset.

Three-dimensional point cloud data are widely used in many fields, as they can be easily obtained and contain rich semantic information. Recently, weakly supervised segmentation has attracted lots of attention, because it only requires very few labels, thus reducing time-consuming and expensive data annotation efforts for huge amounts of point cloud data. The existing approaches typically adopt softmax scores from the last layer as the confidence for selecting high-confident point predictions. However, such approaches can ignore the potential value of a large number of low-confidence point predictions under traditional metrics. In this work, we propose an uncertainty-guided contrastive learning (UCL) framework for weakly supervised point cloud segmentation. A novel uncertainty metric based on prototype entropy (PE) is presented to estimate the reliability of model predictions. With this metric, we propose a negative contrastive learning module exploiting negative pseudo-labels of predictions with low reliability and an active contrastive learning module enhancing feature learning of segmentation models by predictions with high reliability. We also propose a generic multiscale feature perturbation method to expand a wider perturbation space. Extensive experimental results on indoor and outdoor point cloud datasets demonstrate that the proposed method achieves competitive performance.

This article is concerned with distributed resilient load frequency control (LFC) for multi-area power interconnection systems against jamming attacks. First, considering uncertainties and high dimension nonlinearity, the model-free adaptive control (MFAC) model is adopted for the power system, in which only input and output (I/O) data are used. Second, jamming attacks are modeled in a stochastic process, and a multistep predictive compensation algorithm is developed to mitigate the impact of jamming attacks. Then, the distributed MFAC protocol with predictive compensation algorithm is designed such that the frequency tracking errors under the predictive compensation algorithm of multi-area power interconnection systems converge consensually into a small neighborhood of origin in the mean square sense. Simulation results show the effectiveness of the approach.

This brief considers the security control problem for nonlinear cyber-physical systems (CPSs) against jamming attacks. First, a novel event-based model-free adaptive control (MFAC) framework is established. Second, a multistep predictive compensation algorithm (PCA) is developed to make compensation for the lost data caused by jamming attacks, even consecutive attacks. Then, an event-triggering mechanism with the dead-zone operator is introduced in the adaptive controller, which can effectively save communication resources and reduce the calculation burden of the controller without affecting the control performance of systems. Moreover, the boundedness of the tracking error is ensured in the mean-square sense, and only the input/output (I/O) data are used in the whole design process. Finally, simulation comparisons are provided to show the effectiveness of our method.

Extreme learning machine (ELM) owns the advantages of less computational efforts and simple topology with single-hidden layer structure. However, the performance of plain ELM is sensitive to the input weights, bias, and the number of hidden neurons; and the former two are randomly generated. This paper develops a restricted Boltzmann strategy combined with Moore-Penrose generalized inverse to learn topological parameters in both input and output layers. A novel extreme learning model based on the restricted Boltzmann ELM, constructs a feature mapping and recursively tune the weights between input neurons and hidden neurons. The contribution of this paper is to provide a simple ELM topological network to handle low dimensionality problem with the merit of better accuracy and stability. The proposed methodology is evaluated on University of California Irvine (UCI) benchmark datasets for classification issue, and then extended to gas path fault diagnosis for a turbofan engine. The experimental results confirm the superiority to plain ELM.