The vulnerability of graph-based recommender systems to spurious correlations has become a significant obstacle to their practical deployment, hindering their robustness in out-of-distribution (OOD) scenarios. While existing approaches offer partial solutions, they are limited by fundamental shortcomings: model-centric approaches reliant on predefined causal graphs often suffer from suboptimal performance due to complex and dynamic environmental influences. These methods typically require identifying an environmental label or performing feature decoupling, but hidden environments are often difficult to model. Furthermore, existing general feature decoupling methods fail to account for the unique structural characteristics of graphs. To overcome these challenges, we advocate for a shift towards explicit, geometrically-grounded disentanglement. Hyperbolic geometry is particularly suited for this task due to its capacity to model the inherent hierarchies of user interests. We introduce C-HyPOD : Causal Hyperbolic Representation Learning with Prototype Orthogonal Disentanglement, a novel framework designed for graph-based OOD recommendation. Unlike traditional methods, C-HyPOD transforms disentanglement into a concrete geometric task. It introduces a global interest space by learning a single set of universal interest prototypes. They provides a superior geometric foundation for ensuring these prototypes are well-separated and semantically distinct. To ensure a complete separation and prevent information leakage, a targeted orthogonality constraint is then applied. This constraint purifies the aggregated causal representation by forcing it to be orthogonal to the spurious representation in the tangent space, thereby eliminating their linear correlation. Extensive experiments on four public datasets demonstrate that C-HyPOD significantly improves OOD robustness and recommendation performance, surpassing state-of-the-art methods.

Knowledge Graphs (KGs) enhance recommender systems but face challenges from inherent noise, sparsity, and Euclidean geometry's inadequacy for complex relational structures, critically impairing representation learning, especially for long-tail entities. Existing methods also often lack adaptive multi-source signal fusion tailored to item popularity. This paper introduces SPARK, a novel multi-stage framework systematically tackling these issues. SPARK first employs Tucker low-rank decomposition to denoise KGs and generate robust entity representations. Subsequently, an SVD-initialized hybrid geometric GNN concurrently learns representations in Euclidean and Hyperbolic spaces; the latter is strategically leveraged for its aptitude in modeling hierarchical structures, effectively capturing semantic features of sparse, long-tail items. A core contribution is an item popularity-aware adaptive fusion strategy that dynamically weights signals from collaborative filtering, refined KG embeddings, and diverse geometric spaces for precise modeling of both mainstream and long-tail items. Finally, contrastive learning aligns these multi-source representations. Extensive experiments demonstrate SPARK's significant superiority over state-of-the-art methods, particularly in improving long-tail item recommendation, offering a robust, principled approach to knowledge-enhanced recommendation. Implementation code is anonymously online. https://github.com/Applied-Machine-Learning-Lab/SPARK.

Modern recommendation systems face significant challenges in processing multimodal sequential data, particularly in temporal dynamics modeling and information flow coordination. Traditional approaches struggle with distribution discrepancies between heterogeneous features and noise interference in multimodal signals. We propose FindRec (Flexible unified information disentanglement for multi-modal sequential Rec ommendation), introducing a novel ''information flow-control-output'' paradigm. The framework features two key innovations: (1) A Stein kernel-based Integrated Information Coordination Module (IICM) that theoretically guarantees distribution consistency between multimodal features and ID streams, and (2) A cross-modal expert routing mechanism that adaptively filters and combines multimodal features based on their contextual relevance. Our approach leverages multi-head subspace decomposition for routing stability and RBF-Stein gradient for unbiased distribution alignment, enhanced by linear-complexity Mamba layers for efficient temporal modeling. Extensive experiments on three real-world datasets demonstrate FindRec's superior performance over state-of-the-art baselines, particularly in handling long sequences and noisy multimodal inputs. Our framework achieves both improved recommendation accuracy and enhanced model interpretability through its modular design. The implementation code is available anonymously online for easy reproducibility https://github.com/Applied-Machine-Learning-Lab/FindRec.

Diffusion models (DMs) have emerged as the new state-of-the-art family of deep generative models. To gain deeper insights into the limitations of diffusion models in recommender systems, we investigate the fundamental structural disparities between images and items. Consequently, items often exhibit distinct anisotropic and directional structures that are less prevalent in images. However, the traditional forward diffusion process continuously adds isotropic Gaussian noise, causing anisotropic signals to degrade into noise, which impairs the semantically meaningful representations in recommender systems.

Fall, a leading cause of accidental death and injury in older adults aged 65 and above, has become a rapidly growing health concern in aging populations worldwide. Data-driven methods integrating depth imaging technology have received growing attention in automated fall risk assessment owing to their noninvasiveness and less dependence on healthcare professionals. However, most existing depth image data-based models neglect the inherent physiological and potential functional connections and lack sufficient real-world data validation. To fill the research gap, we developed a novel approach named multiscale skeletal transformer (MSS-Former), leveraging depth image technology and deep-learning models for effective fall risk prediction. Our contributions mainly consist of four parts. First, we introduced a multimodel output feature fusion transformer in fall risk prediction, enabling output merging and weighting from multiple model streams dynamically. Second, we developed an innovative scheme to construct interjoint skeletal topology, systematically focusing on joints’ intrinsic physiological and potential functional connections. Third, we constructed a ResNet-FPN, greatly enhancing multiscale feature extraction capabilities. Fourth, we conducted a field study in a local hospital and performed a comprehensive validation of our developed approach. The comparison results show that our approach achieved outstanding predictive performance, surpassing state-of-the-art methods on the real-world data set, with accuracy, precision, recall, and F1 scores of 97.84%, 97.33%, 96.97%, and 96.92%, respectively. In practice, the proposed approach would be of great value in the timely identification for individuals at high fall risk and facilitate decision making to take appropriate interventions.

At present, the application of Electroencephalogram (EEG) signal classification to human intention-behavior prediction has become a hot topic in the brain computer interface (BCI) research field. In recent studies, the introduction of convolutional neural networks (CNN) has contributed to substantial improvements in the EEG signal classification performance. However, there is still a key challenge with the existing CNN-based EEG signal classification methods, the accuracy of them is not very satisfying. This is because most of the existing methods only utilize the feature maps in the last layer of CNN for EEG signal classification, which might miss some local and detailed information for accurate classification. To address this challenge, this paper proposes a multi-scale CNN model-based EEG signal classification method. In this method, first, the EEG signals are preprocessed and converted to time-frequency images using the short-time Fourier Transform (STFT) technique. Then, a multi-scale CNN model is designed for EEG signal classification, which takes the converted time-frequency image as the input. Especially, in the designed multi-scale CNN model, both the local and global information is taken into consideration. The performance of the proposed method is verified on the benchmark data set 2b used in the BCI contest IV. The experimental results show that the average accuracy of the proposed method is 73.9 percent, which improves the classification accuracy of 10.4, 5.5, 16.2 percent compared with the traditional methods including artificial neural network, support vector machine, and stacked auto-encoder.