Heterogeneous graph neural networks (HGNNs) have emerged as a powerful approach for modeling heterogeneous graphs. However, graphs often suffer from missing attributes, which poses significant challenges for HGNNs. Existing methods focus on attribute completion while neglecting the impact of class imbalance, limiting their effectiveness. To address attribute-missing heterogeneous graphs with class imbalance, we propose a novel attribute completion and class-balanced framework for HGNNs. Specifically, to mitigate class imbalance, we design a high-order augmented class-balancing module to reduce representation bias. In addition, to complete missing attributes, we propose a self-supervised heterogeneous attribute completion mechanism. To our knowledge, this is the first work focusing on class imbalance in heterogeneous graphs with missing attributes. On three benchmark datasets, our framework surpasses the state of the art attribute-missing and class-imbalance models and achieves average improvements of 2.45% in node classification and 5.17% in node clustering, respectively. Furthermore, our high-order augmented class-balancing module can be employed as a plug-in component. Its integration with baselines improves the F1 score by 0.44% on average. The code and data of this work are available at https://github.com/cherryyyy5/GraphHOAC.git.
Grouped prediction over semantically labeled criteria may require signed directional constraints and non-additive interaction modeling. We study these requirements jointly through typed signed-monotone dual Choquet-Stieltjes networks (TSM-DCSNs), a constrained family for grouped inputs with adverse, benefit, and neutral score semantics. For a fixed typed score map, a shared threshold law drives trigger and veto Choquet-Stieltjes branches. Anchored additivity characterizes conditions that force affine collapse. Every normalized piecewise-linear threshold law on a fixed knot grid, including the affine law, is finitely identifiable on a nondegenerate criterion ray. Designated probes permit exact local recovery of branch weights and supported dual-capacity values, with stable recovery under bounded perturbations. For fixed k, block-k-additive instances admit polynomial-time forward evaluation and supported Möbius inversion from a polynomial number of designated probes. Synthetic probe experiments confirm machine-precision recovery in the matched noiseless regime and monotone error growth under bounded noise. On five grouped tabular benchmarks, all structured configurations attain zero signed-direction violations; validation-based rankings favor the full member on some tasks and reduced members on others. Because within-group score constructions differ, these comparisons are not single-component ablations. An outer-test-isolated nested pilot gate matches the descriptive full-versus-additive decision on only two of five datasets and makes three false skips relative to the stated descriptive reference, so it is not validated for pre-full-model screening. The contribution is a validation-selectable constrained family for semantically specified grouped prediction with analyzable threshold geometry and probe-based recovery of interaction structure, not a claim of uniform predictive superiority.
Accurate high-resolution spatio-temporal air quality inference based on sparse monitoring stations is essential for environmental governance and public health. However, prevailing deep learning approaches often treat inference as direct regression on observed air quality concentrations, neglecting the distinct effects of a long-term static baseline and short-term dynamic perturbations on the inference. This limits the model’s generalization ability and may cause spatial bias in the inference. In this paper, we propose a novel Spatio-Temporal Air Quality Inference Network (SAQIN) model, which provides an explicit decomposition of the static and dynamic effects in the inference. The static branch integrates high-resolution semantic segmentation, global context from remote sensing, population density, and elevation to construct a high-fidelity environmental prior. The dynamic branch employs multi-head self-attention mechanism to jointly encode geographic relationships and inter-pollutant chemical coupling effects. We further develop a residual learning module to model the inference process as a residual correction anchored to multiple reference stations and aggregates predictions through stepwise multi-station fusion with distance-weighted averaging, which successfully eliminates the Voronoi-partition-induced discontinuities and yielding a globally smooth physically plausible air quality inference field. SAQIN has been shown to demonstrate superior accuracy and strong zero-shot cross-domain generalisation when evaluated on large-scale real-world datasets, thus outperforming state-of-the-art methods across a range of criteria pollutants.
Gaussian processes (GPs) are widely used for modeling dynamical systems due to their ability to provide principled uncertainty estimation and to incorporate prior knowledge via the mean or kernel function. However, GPs typically exhibit poor extrapolation performance outside the training region. To address this limitation while preserving the favorable uncertainty behavior of standard GPs, we propose a novel kernel design that integrates prior models-either analytical physical models or imperfect simulators-into the kernel function of the GP. Our approach is motivated by the assumption that similarity in the prior model implies similarity in the true system dynamics. Specifically, we transform the input space using the prior model and apply a base kernel (e.g., Radial Basis Function or Matérn) to construct a prior-model-informed kernel that reflects this assumption. To further enhance modeling flexibility, we add a standard residual kernel to correct discrepancies between the prior model and the true system. This yields our final model: a GP with prior-model-informed kernel (GP-PI-K). Through extensive experiments on benchmark dynamical systems, we demonstrate that GP-PI-K consistently outperforms existing baselines, including standard GPs, GPs with physics-informed or neural network mean functions, and deep kernel learning models, in terms of one-step and multi-step prediction accuracy, uncertainty estimation, and uncertainty calibration. Moreover, GP-PI-K achieves superior performance in downstream applications such as active learning and model-based reinforcement learning.
Rationalization models have recently garnered significant attention for enhancing the interpretability of natural language processing by first using a generator to select the most relevant pieces from the text with respect to the label, before passing the text input to the predictor. However, the robustness of the rationalization models is not sufficiently investigated. Specifically, this paper explores the robustness of rationalization models against backdoor attacks, which has been ignored by previous studies. Surprisingly, we find that conventional backdoor attack techniques fail to inject triggers into the rationalization model because its generator can filter out bad triggers. Considering this, we further propose a novel backdoor attack method named as BadRNL designed specially for the rationalization models. The core idea of BadRNL is first to search for the personalized trigger for each specific dataset and then manipulate the rationales and labels to conduct attacks. Besides, BadRNL controls the order of sample learning through poison-priority sampling strategies. Experimental results across five diverse NLP datasets—FEVER, MultiRC, Beer, Hotel, and Movie—demonstrate that BadRNL consistently achieves near 100% attack success rates (ASR). Crucially, the method maintains high classification accuracy and rationale quality on benign samples, highlighting a significant security risk for interpretable NLP systems deployed in real-world supply-chain scenarios.
