Hate Video Detection (HVD) is crucial for online ecosystems. Existing methods assume identical distributions between training (source) and inference (target) data. However, hateful content often evolves into irregular and ambiguous forms to evade censorship, resulting in substantial semantic drift and rendering previously trained models ineffective. Test-Time Adaptation (TTA) offers a solution by adapting models during inference to narrow the cross-domain gap, while conventional TTA methods target mild distribution shifts and struggle with the severe semantic drift in HVD. To tackle these challenges, we propose SCANNER, the first TTA framework tailored for HVD. Motivated by the insight that, despite the evolving nature of hateful manifestations, their underlying cores remain largely invariant (i.e., targeting is still based on characteristics like gender, race, etc), we leverage these stable cores as a bridge to connect the source and target domains. Specifically, SCANNER initially reveals the stable cores from the ambiguous layout in evolving hateful content via a principled centroid-guided alignment mechanism. To alleviate the impact of outlier-like samples that are weakly correlated with centroids during the alignment process, SCANNER enhances the prior by incorporating a sample-level adaptive centroid alignment strategy, promoting more stable adaptation. Furthermore, to mitigate semantic collapse from overly uniform outputs within clusters, SCANNER introduces an intra-cluster diversity regularization that encourages the cluster-wise semantic richness. Experiments show that SCANNER outperforms all baselines, with an average gain of 4.69% in Macro-F1 over the best.

Weight Averaging (WA) has emerged as a powerful technique for enhancing generalization by promoting convergence to a flat loss landscape, which correlates with stronger out-of-distribution performance. However, applying WA directly to multi-modal domain generalization (MMDG) is challenging: differences in optimization speed across modalities lead WA to overfit to faster-converging ones in early stages, suppressing the contribution of slower yet complementary modalities, thereby hindering effective modality fusion and skewing the loss surface toward sharper, less generalizable minima. To address this issue, we propose MBCD, a unified collaborative distillation framework that retains WA's flatness-inducing advantages while overcoming its shortcomings in multi-modal contexts. MBCD begins with adaptive modality dropout in the student model to curb early-stage bias toward dominant modalities. A gradient consistency constraint then aligns learning signals between uni-modal branches and the fused representation, encouraging coordinated and smoother optimization. Finally, a WA-based teacher conducts cross-modal distillation by transferring fused knowledge to each uni-modal branch, which strengthens cross-modal interactions and steer convergence toward flatter solutions. Extensive experiments on MMDG benchmarks show that MBCD consistently outperforms existing methods, achieving superior accuracy and robustness across diverse unseen domains.

Network traffic anomaly detection represents a critical cybersecurity task, yet widespread encryption makes this task increasingly challenging. In response, image-based methods that model traffic as visual patterns have emerged as the dominant approach. However, this work pioneers the identification of a pervasive "full-frequency"characteristic and an associated limitation termed "spectral mismatch"within this paradigm. Specifically, while encrypted traffic exhibits prominent high-frequency components, mainstream reconstruction methods demonstrate an inherent bias toward learning low-frequency information. This fundamental mismatch results in incomplete representations that consequently degrade anomaly detection performance. To address this challenge, we propose FreeUp, a novel frequency-decoupled framework designed explicitly for encrypted traffic analysis. FreeUp decomposes traffic data into distinct low- and high-frequency bands, processing them through separate, dedicated branches along with a customized training strategy that ensures stable and independent frequency-specific learning. Furthermore, recognizing that simple reconstruction error proves inadequate for evaluating dual-branch architectures, we introduce an uncertainty-inspired fusion scoring mechanism. This mechanism quantifies the reconstruction uncertainty of the frequency-specific branches and dynamically integrates their outputs, yielding a more comprehensive and reliable anomaly score. Extensive experiments across multiple benchmarks demonstrate that FreeUp consistently outperforms state-of-the-art baselines. The code is available at https://github.com/ikun0124/FreeUp.

Multimodal Large Language Model (MLLM) Personalization is a critical research problem that facilitates personalized dialogues with MLLMs targeting specific entities (known as personalized concepts). However, existing methods and benchmarks focus on the simple, context-agnostic visual identification and textual replacement of the personalized concept (e.g., "A yellow puppy" -> "Your puppy Mochi"), overlooking the ability to support long-context conversations. An ideal personalized MLLM assistant is capable of engaging in long-context dialogues with humans and continually improving its experience quality by learning from past dialogue histories. To bridge this gap, we propose LCMP, the first Long-Context MLLM Personalization evaluation benchmark. LCMP assesses the capability of MLLMs in perceiving variations of personalized concepts and generating contextually appropriate personalized responses that reflect these variations. As a strong baseline for LCMP, we introduce a novel training-free and state-aware framework TAME. TAME endows MLLMs with double memories to manage the temporal and persistent variations of each personalized concept in a differentiated manner. In addition, TAME incorporates a new training-free Retrieve-then-Align Augmented Generation (RA2G) paradigm. RA2G introduces an alignment step to extract the contextually fitted information from the multi-memory retrieved knowledge to the current questions, enabling better interactions for complex real-world user queries. Experiments on LCMP demonstrate that TAME achieves the best performance, showcasing remarkable and evolving interaction experiences in long-context scenarios.

The proliferation of harmful memes on online media poses significant risks to public health and stability. Existing detection methods heavily rely on large-scale labeled data for training, which necessitates substantial manual annotation efforts and limits their adaptability to the continually evolving nature of harmful content. To address these challenges, we present ALARM, the first lAbeL-free hARmful Meme detection framework powered by Large Multimodal Model (LMM) agent self-improvement. The core innovation of ALARM lies in exploiting the expressive information from "shallow" memes to iteratively enhance its ability to tackle more complex and subtle ones. ALARM consists of a novel Confidence-based Explicit Meme Identification mechanism that isolates the explicit memes from the original dataset and assigns them pseudo-labels. Besides, a new Pairwise Learning Guided Agent Self-Improvement paradigm is introduced, where the explicit memes are reorganized into contrastive pairs (positive vs. negative) to refine a learner LMM agent. This agent autonomously derives high-level detection cues from these pairs, which in turn empower the agent itself to handle complex and challenging memes effectively. Experiments on three diverse datasets demonstrate the superior performance and strong adaptability of ALARM to newly evolved memes. Notably, our method even outperforms label-driven methods. These results highlight the potential of label-free frameworks as a scalable and promising solution for adapting to novel forms and topics of harmful memes in dynamic online environments.

LiDAR sensors are critical for 3D scene perception in autonomous driving, providing high spatial accuracy but suffering from limited temporal resolution. Point cloud interpolation, which reconstructs intermediate frames from sparsely captured inputs, effectively enhances temporal continuity. However, the challenges posed by non-linear and non-rigid motion in dynamic scenarios remain unresolved by existing methods, which often sacrifice accuracy for efficiency. To solve the above problems, we propose MSTSINet (Multiscale Transformer and SSM Interpolation Network), a novel framework that combines multiscale forward flow prediction with non-linear trajectory compensation for fast and accurate interpolation. Unlike conventional bidirectional flow-based methods, our forward flow-guided approach reduces inference time while improving robustness to complex motion. MSTSINet integrates multiscale motion modeling through a novel hierarchical feature extraction and semantic-aware cross-frame attention, progressively improving forward flow estimation for accurate non-linear trajectory prediction. To further improve motion fidelity and temporal coherence, we introduce a point-wise compensation mechanism during inference and design a novel multiscale loss function that supervises forward flow learning across varying temporal resolutions. Experiments on autonomous driving benchmarks demonstrate that MSTSINet significantly outperforms existing state-of-the-art methods, achieving up to a 26.39% reduction in Chamfer Distance and substantially improving inference speed. These advantages make it well-suited for real-time deployment in autonomous driving scenarios.

Ensuring reliable object detection in adverse conditions is paramount for safe autonomous driving. While cameras and LiDAR struggle in such scenarios, Frequency Modulated Continuous Wave (FMCW) radar offers superior robustness, but its low azimuth resolution can hinder object distinction. Leveraging temporal information across multiple frames can mitigate this limitation, but the radar's inherent frame rate poses a constraint. This paper proposes a deep learning-based interpolate-then-detect framework for FMCW radar. In this framework, the radar frame sequence is first interpolated to synthesize high-quality intermediate frames, and object detection is subsequently performed on the resulting densified sequence, thereby improving detection accuracy and temporal stability. Specifically, we propose a unified network that employs multiple Gaussian Splatting Efficient Attention (GSEA) modules. Each GSEA module combines 3D Gaussian splatting to provide a probabilistic, motion-aware embedding that captures spatial uncertainty and variability in radar measurements, while split multi-head attention enlarges the effective spatio-temporal receptive field and selectively emphasizes the most informative features across frames. Experimental results demonstrate that the proposed approach achieves up to a 9.6 % improvement in average precision and exhibits enhanced robustness in noisy environments, highlighting its potential for accurate and reliable object detection in challenging driving scenarios.

Micro-video popularity prediction (MVPP) plays a crucial role in various downstream applications. Recently, multimodal methods that integrate multiple modalities to predict the popularity have exhibited impressive performance. However, these methods face several unresolved issues: (1) limited contextual information and (2) incomplete modal semantics. Incorporating relevant videos and performing full fine-tuning on pre-trained models typically achieves powerful capabilities in addressing these issues. However, this paradigm is not optimal due to its weak transferability and scarce downstream data. Inspired by prompt learning, we propose ICPF, a novel In-Context Prompt-augmented Framework to enhance popularity prediction. ICPF maintains a model-agnostic design, facilitating seamless integration with various multimodal fusion models. Specifically, the multi-branch retriever first retrieves similar modal content through within-modality similarities. Next, in-context prompt generator extracts semantic prior features from retrieved videos and generates in-context prompts, enriching pre-trained models with valuable contextual knowledge. Finally, knowledge-augmented predictor captures complementary features including modal semantics and popularity information. Extensive experiments conducted on three real-world datasets demonstrate the superiority of ICPF compared to 14 competitive baselines.

Multimodal learning with incomplete modality is practical and challenging. Recently, researchers have focused on enhancing the robustness of pre-trained MultiModal Transformers (MMTs) under missing modality conditions by applying learnable prompts. However, these prompt-based methods face several limitations: (1) incomplete modalities provide restricted modal cues for task-specific inference, (2) dummy imputation for missing content causes information loss and introduces noise, and (3) static prompts are instance-agnostic, offering limited knowledge for instances with various missing conditions. To address these issues, we propose RAGPT, a novel Retrieval-AuGmented dynamic Prompt Tuning framework. RAGPT comprises three modules: (I) the multi-channel retriever, which identifies similar instances through a within-modality retrieval strategy, (II) the missing modality generator, which recovers missing information using retrieved contexts, and (III) the context-aware prompter, which captures contextual knowledge from relevant instances and generates dynamic prompts to largely enhance the MMT’s robustness. Extensive experiments conducted on three real-world datasets show that RAGPT consistently outperforms all competitive baselines in handling incomplete modality problems.

Predicting user engagement on social media platforms is increasingly vital for applications such as recommendations, Internet marketing, and content moderation. Existing methods face challenges in effectively capturing contextual information and ensuring relevance, often leading to the retrieval of semantically similar but contextually irrelevant UGCs. To address these challenges, we introduce DRUEP, a distilled retrieval-augmented framework that explores distilled multimodal representations through the information bottleneck technique to enhance UGC relevance retrieval, improving the quality of retrieved UGC for more accurate user engagement prediction. Our approach is able to effectively reduce redundant multi-modal noise, optimizes retrieval efficiency, and captures dynamic nature of UGC. Extensive experiments on multimodal large-scale datasets demonstrate that DRUEP significantly outperforms existing state-of-the-art methods.

With the rapid expansion of the internet, IP geolocation has become crucial for network security, content delivery, and compliance. However, existing methods struggle with dynamic, noisy networks, especially in handling topology changes, noisy data, and efficient node selection. To address these, we introduce PINGeo, a novel framework based on graph convolutional networks. PINGeo combines two innovations: (1) adaptive graph pruning using node importance metrics to retain critical nodes, optimizing both efficiency and accuracy, and (2) Gaussian noise perturbation to enhance robustness by simulating real-world fluctuations. By applying these methods, PINGeo improves geolocation accuracy and model robustness in noisy environments. Experimental results on public datasets show that PINGeo outperforms state-of-the-art methods in accuracy and robustness, offering a promising solution for robust IP geolocation.

Accurately segmenting polyp regions in colonoscopy images is crucial for the diagnosis and intervention of colorectal cancer. However, the task of polyp segmentation remains challenging due to the diverse size and shape variations among polyps, their extreme similarity to the background, and frequent rotation of the lens, which further increases the diversity in polyp presentation. To address these challenges effectively, we propose a comprehensive polyp segmentation network (CPSNet). Specifically, we introduce a Comprehensive Spatial Feature Extraction Module (CFEM) that progressively and densely integrates features while forming receptive windows with various shapes. This enables enhanced perception of polyps at manifold sizes and shapes. Additionally, we propose a Fine-grained Region Strengthen Module (FGSM) to supplement uncertain areas around polyps by mitigating background noise interference. In terms of training strategy, we further introduce a Rotation-augmented Constrained Loss (RC Loss), which reinforces consistency constraints on polyp images under multiple rotation angles. Qualitative and quantitative experiments conducted on five public datasets demonstrate both the plug-and-play capability of CFEM as well as the effectiveness and excellence achieved by CPSNet.

Online crowdfunding platforms have been gaining increasing popularity due to their convenience in soliciting social capital from the public. These platforms offer valuable opportunities for fundraisers to bring their creative products to life and support pro-social projects. However, the relatively low success rate of crowdfunding campaigns highlights the need for better strategies. While existing studies have explored various factors that contribute to crowdfunding success, they often overlook the intricate relationships between different aspects of crowdfunding. In this paper, we propose a novel model called Commonality Augmented Multimodal Disentanglement (CAMD) for predicting crowdfunding success. It can disentangle the roles of different data modalities, separating common factors from specific ones. Besides, we enhance the disentangled commonality using an augmentation network to achieve balanced representation of the skewed interrelations between different modalities. At last, we introduce a cross-attention-based multimodal fusion mechanism that further improves model performance by highlighting the crucial role of crowdfunding attributes. Experiments conducted on two large-scale crowdfunding datasets demonstrate the effectiveness and generalizability of our model.

Unsupervised graph anomaly detection aims to identify irregular patterns in graph-structured data without relying on labeled anomalies. Graph neural networks (GNNs) have advanced GAD by learning effective graph representations through neighborhood aggregation. However, challenges such as anomalous information diffusion impede GNNs to accurately distinguish anomalies from normal ones. To bridge this gap, we propose a novel framework named Selective Anomalous information Filtering for Enhance unsupervised graph anomaly detection (SAFE). SAFE allows nodes to selectively filter anomalous information, preventing the spread of anomalous noise to normal nodes while allowing anomalies to assimilate features from their neighbors. This strategy enhances the reconstruction error disparity between normal and anomalous nodes, thereby improving the accuracy of anomaly detection. Extensive experiments on both synthetic and real-world datasets demonstrate the significant performance gains of SAFE over existing methods.

Diffusion models have shown great potential for super-resolution by effectively mapping high-resolution images from low-resolution inputs. However, the heavy inference cost remains a significant challenge. In this work, we propose Efficient self-correcting Diffusion Bridge (EDB), a novel framework for diffusion bridge-based super-resolution. EDB introduces two key components: non-Markovian implicit sampling to improve sampling efficiency and self-correcting training to reduce estimation error. Applied to image super-resolution (SR), EDB effectively balances enhanced sampling efficiency with high image quality. Extensive quantitative and qualitative comparisons with advanced methods reveal that our method achieves competitive or even superior super-resolution performance metrics with higher sampling efficiency, highlighting its capability for accurate and efficient super-resolution of degraded images.

Instead of using a standard diffusion model, most diffusion-based speech enhancement methods additionally incorporate a mean interpolation strategy into the diffusion process. However, the role and impact of this strategy remain unclear. In this study, we investigate mean interpolation from a data augmentation perspective, demonstrating that it serves as a specific form of distribution augmentation and provides a unified formula for mean interpolation, which encompasses current variations of such strategy. Building on this insight, we propose DANS (Distribution Augmentation with Noise Shuffling), an advanced distribution augmentation method that further expands the training distribution through noise-shuffling techniques. Experimental results show that DANS consistently outperforms existing methods in matched, cross-dataset, and low-data scenarios. Codes are publicly available at https://github.com/ICDM-UESTC/DANCE.

Multi-modal knowledge graph (MMKG) completion aims to infer missing links in knowledge graphs by leveraging both structural information and multi-modal data, such as images and text. While existing methods focus on multi-modal feature fusion, they often fail to disentangle modality-specific and shared features, leading to redundant and suboptimal embeddings. Furthermore, implicit semantic correlations between entities are frequently overlooked. To address these issues, we propose INCADE, an Implicit Entity Community-Aware Disentangled Expert framework that jointly captures implicit high-order correlations among entities and disentangles modality-specific and shared representations. INCADE introduces an entity community-aware encoder to model implicit entity interactions and auxiliary tasks (e.g., modality memorization and unlearning) to enforce disentanglement. This advantage enables the model to easily utilizes both explicit and implicit multi-modal information. Extensive experiments on standard MMKG benchmarks demonstrate that INCADE achieves state-of-the-art performance, outperforming 14 baseline models. Code is available at https://github.com/shchli/INCADE.

Detecting fake news videos has emerged as a critical task due to their profound implications in politics, finance, and public health. However, existing methods often fail to distinguish real videos from their subtly manipulated counterparts, resulting in suboptimal performance. To address this limitation, we propose REAL, a novel model-agnostic REtrieval-Augmented prototype-aLignment framework. REAL first introduces an LLM-driven video retriever to identify contextually relevant samples for a given target video. Subsequently, a dual-prototype aligner is carefully developed to model two distinct prototypes: one representing authentic patterns from retrieved real news videos and the other encapsulating manipulation-specific patterns from fake samples. By aligning the target video’s representations with its ground-truth prototype while distancing them from the opposing prototype, the aligner captures manipulation-aware representations capable of detecting even subtle video manipulations. Finally, these enriched representations are seamlessly integrated into existing detection models in a plug-and-play manner. Extensive experiments on three benchmarks demonstrate that REAL largely enhances the detection ability of existing methods. The code and data for reproducing the results are available at https://github.com/Jian-Lang/REAL.

Micro-video popularity prediction (MVPP) plays a crucial role in diverse real-world applications. While recent popularity prediction methods leveraging multimodal content have shown impressive performance, they face two unresolved challenges: (1) neglecting the presence of noise in micro-videos and (2) failing to address incomplete modalities. To tackle these issues, we propose FMOE, a novel Flexible Mixture of Modality Experts framework designed for robust MVPP. FMOE employs a two-stage learning strategy, consisting of pre-training and post-training phases. Specifically, during pre-training, FMOE extracts discriminative unimodal representations by training modality-specific experts. In the post-training phase, FMOE dynamically captures cross-modal correlations by mixing experts. Subsequently, FMOE quantifies uncertainty within micro-videos using an uncertainty-aware router, which maps modal representations into a Gaussian distribution space. Extensive experiments on three real-world datasets demonstrate that FMOE significantly outperforms all competitive baselines by a large margin.

Multimodal sentiment analysis is critical in many real-world applications like smart cities, healthcare, and human-computer interaction, where sentiment is conveyed through various modalities, including text, audio, and video. However, existing methods still face several critical challenges in mitigating the impact of random modality loss, particularly in preserving the reliability of emotional patterns. To address these limitations, we introduce UniMSA, a novel framework for multimodal sentiment analysis with missing modalities. It addresses the missing data issue by improving the bipolar emotional uncertainty learning. Our approach enhances the reliability of sentiment analysis by integrating both positive and negative emotional uncertainty estimations to recover emotional patterns in randomly missing modalities. Extensive experiments conducted on large-scale multimodal sentiment datasets demonstrate the effectiveness of UniMSA in comparison to state-of-the-art methods.

This work presents a systematic investigation into the trustworthiness of explanations generated by self-interpretable graph neural networks (GNNs), revealing why models trained with different random seeds yield inconsistent explanations. We identify redundancy—resulting from weak conciseness constraints—as the root cause of both explanation inconsistency and its associated inaccuracy, ultimately hindering user trust and limiting GNN deployment in high-stakes applications. Our analysis demonstrates that redundancy is difficult to eliminate; however, a simple ensemble strategy can mitigate its detrimental effects. We validate our findings through extensive experiments across diverse datasets, model architectures, and self-interpretable GNN frameworks, providing a benchmark to guide future research on addressing redundancy and advancing GNN deployment in critical domains. Our code is available at \url{https://github.com/ICDM-UESTC/TrustworthyExplanation}.

Micro-video popularity prediction (MVPP) plays a crucial role in numerous real-world applications, including product marketing and recommendation systems. While existing methodologies predominantly assume complete modalities during multimodal learning, this assumption often fails to hold in practical scenarios due to various constraints, such as privacy concerns or data integrity issues. To address this limitation, we propose SCRAG, a novel Self-Correlation Retrieval-Augmented Generative framework designed to enhance missing-modality robustness in MVPP. SCRAG operates in a retrieval-guided generation manner that explores relevant knowledge to enhance the reconstruction of missing content, which consists of two primary components: (1) a self-correlation retriever and (2) a multimodal mixture-of-experts generator. It first acquires instances pertinent to the missing content through multimodal prompt alignment. Subsequently, the generator extracts contextual modal information from the retrieved context-rich instances. By learning the joint distribution of modalities, SCRAG effectively recovers missing content and addresses the modal heterogeneity challenge inherent in cross-modal generation approaches. Extensive experiments conducted on three real-world datasets demonstrate that SCRAG consistently outperforms state-of-the-art baselines, underscoring its effectiveness in handling incomplete modalities and improving the accuracy of micro-video popularity prediction.

Zero-shot composed image retrieval (ZS-CIR) is a challenging task that aims to retrieve images similar to a composed query of a reference image and a description, without relying on training on triplet datasets. Existing methods for this task often rely on predefined, fixed retrieval processes that combine the image and the modified text through hand-crafted templates, which suffer from two main issues: non-adaptive retrieval queries and user-unfriendly retrieval processes. To address these limitations, we propose a novel framework - Automatic Multi-Agent Collaboration for Zero-Shot Composed Image Retrieval (AutoCIR). AutoCIR consists of three training-free agents - a planner, a retriever, and a corrector - that work together to iteratively identify and rectify mismatches. The planner guides the retriever by generating a customized target caption for the composed query and further refines this caption to resolve any semantic discrepancies based on feedback. The corrector, equipped with a chain-of-thought reasoning mechanism, conducts an in-depth evaluation of the retrieved results and generates appropriate self-correction actions. Extensive experiments on three benchmarks demonstrate that AutoCIR consistently outperforms previous competitive methods for ZS-CIR.

Traditional multimodal learning approaches often assume that all modalities are available during both the training and inference phases. However, this assumption is often impractical in real-world scenarios due to challenges such as sensor failures, data corruption, or privacy concerns. While recent efforts focus on enhancing the robustness of pre-trained Multimodal Transformers (MTs) under missing modality conditions, mainstream work in this field often overlook reconstructing the missing modalities and rely on static, sample-agnostic prompt-tuning techniques, undermining their efficacy in severe modality missing scenarios. To address these limitations, we propose REDEEM, a novel REtrieval-guiDEd conditional gEnerative fraMework that largely alleviates the modality missing problems on pre-trained MTs. REDEEM consists of a new adaptive retrieval mechanism to identify relevant instances for both modality-complete and -incomplete samples. It then conditions on the remaining modalities and utilizes the retrieved data as experts to effectively recover the missing ones in modality-incomplete instances through a within-modal reconstruction manner. Finally, REDEEM generates sample-aware inter-modal prompts from the retrieved instances to guide MTs in tackling severe modality missing challenges. Comprehensive experiments on three diverse multimodal classification benchmarks demonstrate that REDEEM significantly outperforms competitive baselines.

The rapid spread of rumor content on online micro-video platforms poses significant threats to public health and safety. However, existing Micro-Video Rumor Detection (MVRD) methods are generally black-box, which lacks transparency and makes it difficult to understand the reasoning behind classification decisions. In this work, we introduce ExMRD, a novel Explainable Micro-video Rumor Detection framework designed to generate detailed and coherent explanations for enhancing MVRD. Inspired by the powerful reasoning capacity of Chain-of-Thought (CoT), we introduce a novel inference mechanism called R3CoT-- consisting of Refining, Retrieving, and Reasoning on MVRD. This mechanism enables Multimodal Large Language Models (MLLMs) to reorganize the original video content, retrieve domain knowledge related to rumors, and generate explainable conclusions regarding whether the micro-video contains rumor information. Instead of directly fine-tuning MLLMs for MVRD, which is computationally expensive, we propose a Small Language Reviewer (SLReviewer), which distills the outputs of R3CoT guided MLLMs to ensure efficient and reliable predictions. Extensive experiments on three real-world benchmarks demonstrate that ExMRD significantly outperforms competitive baselines while providing high-quality rationales.

Information diffusion prediction is a vital component for a wide range of social applications, including viral marketing identification and personal recommendation. Prior methods primarily focus on learning target user representation by modeling contextual information from the historical retweet user sequence of a single cascade, overlooking the uncertainties that exist in both historical propagation trajectory and future diffusion trends. In this work, we propose DucDiff, a novel dual-consistent diffusion model for enhancing target user representation used for information diffusion prediction. DucDiff harnesses the distribution generation capability of the diffusion model to generate target user representations from a distributional perspective rather than a fixed vector. Specifically, it captures the multi-latent aspects (i.e., uncertainties) of target user representation from historical and future user sequences, respectively, using disentangled dual denoising modules. Additionally, a shared information bottleneck is designed for the cross-distillation of knowledge between the historical and future denoising modules, eliminating the performance gap between training and inference, while ensuring that future information can be implicitly introduced during the inference phase. Extensive experiments conducted on five datasets demonstrate that DucDiff significantly outperforms state-of-the-art baselines.

Information diffusion prediction is a vital component for a wide range of social applications, including viral marketing identification and precise recommendation. Prior methods focus on modeling contextual information from a single cascade, ignoring rich collaborative information behind historical interactions across various cascades and future data within the cascade. Leveraging such interactions can substantially enhance diffusion prediction performance but presents two major challenges: (1) user intents are usually entangled behind historical interactions; and (2) utilizing future data may introduce severe training-inference discrepancies. We present MIM, a novel information diffusion model merging multi-scale interactions for improving user intent learning and behavior retrieval. Specifically, we convert cascades and social relations into multi-channel hypergraphs, where each channel depicts a common fine-grained user intent behind historical interactions across cascades. By aggregating embeddings learned through multiple channels, we obtain comprehensive intent representations. Second, we decouple past- and future-level temporal influences within a cascade via a dual temporal network. Then we implement past-future knowledge transferring to enhance the knowledge learnt from the dual network via hierarchical knowledge distillation. Extensive experiments conducted on four datasets demonstrate that MIM significantly outperforms various benchmarks.

Temporal knowledge graph reasoning is an essential task that holds immense value in diverse real-world applications. Existing studies mainly focus on leveraging structural and sequential dependencies, excelling in tasks like entity and link prediction. However, they confront a notable interpretability gap in their predictions, a pivotal facet for comprehending model behavior. In this study, we propose an innovative method, LSGAT, which not only exhibits remarkable precision in entity predictions but also enhances interpretability by identifying pivotal historical events influencing event predictions. LSGAT enables concise explanations for prediction outcomes, offering valuable insights into the otherwise enigmatic "black box" reasoning process. Through an exploration of the implications of the most influential events, it facilitates a deeper understanding of the underlying mechanisms governing predictions.

Earnings call transcripts hold valuable insights that are vital for investors and analysts when making informed decisions. However, extracting these insights from lengthy and complex transcripts can be a challenging task. The traditional manual examination is not only time-consuming but also prone to errors and biases. Deep learning-based representation learning methods have emerged as promising and automated approaches to tackle this problem. Nevertheless, they may encounter significant challenges, such as the unreliability of the representation encoding process and certain domain-specific requirements in the context of finance. To address these issues, we propose a novel transcript representation learning model. Our model leverages the structural information of transcripts to effectively extract key insights, while endowing model with explainability via variational information bottleneck. Extensive experiments on two downstream financial tasks demonstrate the effectiveness of our approach.

The success of graph neural networks (GNNs) has spurred numerous new works leveraging GNNs for modeling multivariate time series anomaly detection. Despite their achieved performance improvements, most of them only consider static graph to describe the spatial-temporal dependencies between time series. Moreover, existing works neglect the time and scale-changing structures of time series. In this work, we propose MDGAD, a novel multi-scale dynamic graph structure learning approach for time series anomaly detection. We design a multi-scale graph structure learning module that captures the complex correlations among time series, constructing an evolving graph at each scale. Meanwhile, an anomaly detector is used to combine bilateral prediction errors to detect abnormal data. Experiments conducted on two time series datasets demonstrate the effectiveness of MDGAD.

Recently, the field of Speech Enhancement has witnessed the success of diffusion-based generative models. However, these diffusion-based methods used to take multiple iterations to generate high-quality samples, leading to high computational costs and inefficiency. In this paper, we propose SDFEN (Shallow Diffusion for Fast spEech eNhancement), a novel approach for addressing the inefficiency problem while enhancing the quality of generated samples by reducing the iterative steps in the reverse process of diffusion method. Specifically, we introduce the shallow diffusion strategy initiating the reverse process with an adaptive time step to accelerate inference. In addition, a dedicated noisy predictor is further proposed to guide the adaptive selection of time step. Experiment results demonstrate the superiority of the proposed SDFEN in effectiveness and efficiency.

Graph anomaly detection has received remarkable research interests, and various techniques have been employed for enhancing detection performance. However, existing models tend to learn dataset-specific spurious correlations based on statistical associations. A well-trained model might suffer from performance degradation when applied to newly observed nodes with different environments. To handle this situation, we propose CounterFactual Graph Anomaly Detection model, CFGAD. In this model, we design a gradient-based separator to disentangle node features into class features and environment features. Then, we present a weight-varying diffusion model to combine class features and environment features from different nodes to generate counterfactual samples. These counterfactual samples will be adopted to enhance model robustness. Comprehensive experiments demonstrate the effectiveness of our CFGAD.

Accurate IP geolocation is indispensable for location-aware applications. While recent advances based on router-centric IP graphs are considered cutting-edge, one challenge remain: the prevalence of sparse IP graphs (14.24% with fewer than 10 nodes, 9.73% isolated) limits graph learning. To mitigate this issue, we designate the target host as the central node and aggregate multiple last-hop routers to construct the target-centric IP graph, instead of relying solely on the router with the smallest last-hop latency as in previous works. Experiments on three real-world datasets show that our method significantly improves the geolocation accuracy compared to existing baselines.

Information diffusion prediction is a critical task for many social network applications. However, current methods are mainly limited by the following aspects: user relationships behind resharing behaviors are complex and entangled. To address these issues, we propose MHGFormer, a novel multi-channel hypergraph transformer framework, to better decouple complex user relations and obtain fine-grained user representations. First, we employ designed triangular motifs to decouple user relations into three different level hypergraphs. Second, a position-aware hypergraph transformer is used to refine user relation and obtain high-quality user representations. Extensive experiments conducted on two social datasets demonstrate that MHGFormer outperforms state-of-the-art diffusion models across several settings.

Social media popularity prediction of multimodal user-generated content (UGC) is a crucial task for many real-world applications. However, existing efforts are often limited by missing inter-instance correlations and UGC temporal patterns. To address these issues, we propose a novel time-aware hypergraph Transformer framework, THGFormer. It fully represents inter-instance and intra-instance relations by hypergraphs, captures the temporal dependencies with a time encoder, and enhances UGC's representations via a neighborhood knowledge aggregation. Extensive experiments conducted on two real-world datasets demonstrate that THGFormer outperforms state-of-the-art popularity prediction models across several settings.

Accurate IP geolocation is crucial for location-aware applications. While recent advances in router-centric IP graph methods have garnered attention, they face two persistent challenges: (1) the sparsity problem of IP graphs in rural areas and (2) the limited explainability of current IP geolocation systems. To tackle these issues, we present ExGeo, a novel and explainable graph-based approach for IP geolocation. Specifically, we introduce a target-centric IP graph, reducing sparsity and enhancing contextual information utilization. Additionally, we endow the model with explainability through a variational graph information bottleneck strategy. Experiments on three real-world datasets demonstrate significant accuracy and explainability improvements. Source code is released at https://github.com/ICDM-UESTC/ExGeo.

Accurately predicting the popularity of multimodal user-generated content (UGC) is fundamental for many real-world applications such as online advertising and recommendation. Existing approaches generally focus on limited contextual information within individual UGCs, yet overlook the potential benefit of exploiting meaningful knowledge in relevant UGCs. In this work, we propose RAGTrans, an aspect-aware retrieval-augmented multi-modal hypergraph transformer that retrieves pertinent knowledge from a multi-modal memory bank and enhances UGC representations via neighborhood knowledge aggregation on multi-model hypergraphs. In particular, we initially retrieve relevant multimedia instances from a large corpus of UGCs via the aspect information and construct a knowledge-enhanced hypergraph based on retrieved relevant instances. This allows capturing meaningful contextual information across the data. We then design a novel bootstrapping hypergraph transformer on multimodal hypergraphs to strengthen UGC representations across modalities via customizing a propagation algorithm to effectively diffuse information across nodes and edges. Additionally, we propose a user-aware attention-based fusion module to comprise the enriched UGC representations for popularity prediction. Extensive experiments on real-world social media datasets demonstrate that RAGTrans outperforms state-of-the-art popularity prediction models across settings.

Trip recommendation has emerged as a highly sought-after service over the past decade. Although current studies significantly understand human intention consistency, they struggle with undesired repetitive outcomes that need resolution. We make two pivotal discoveries using statistical analyses and experimental designs: (1) The occurrence of repetitions is intricately linked to the models and decoding strategies. (2) During training and decoding, adding perturbations to logits can reduce repetition. Motivated by these observations, we introduce AR-Trip (Anti Repetition for Trip Recommendation), which incorporates a cycle-aware predictor comprising three mechanisms to avoid duplicate Points-of-Interest (POIs) and demonstrates their effectiveness in alleviating repetition. Experiments on four public datasets illustrate that AR-Trip successfully mitigates repetition issues while enhancing precision.

Point-of-Interest (POI) recommendation involves predicting users' next preferred POI and is becoming increasingly significant in location-based social networks. However, users are often reluctant to trust recommended results due to the lack of transparency in these systems. While recent work on explaining recommender systems has gained attention, prevailing methods only provide post-hoc explanations based on results or rudimentary explanations according to attention scores. Such limitations hinder reliability and applicability in risk-sensitive scenarios. Inspired by the information theory, we propose a self-explainable framework with an ante-hoc view called \M~for next POI recommendation aimed at overcoming these limitations. Specifically, we endow self-explainability to POI recommender systems through compact representation learning using a variational information bottleneck approach. The learned representation further improves accuracy by reducing redundancy behind massive spatial-temporal trajectories, which, in turn, boosts the recommendation performance. Experiments on three real-world datasets show significant improvements in both model explainability and recommendation performance.

Accurately predicting the popularity of micro-videos is crucial for real-world applications such as recommender systems and identifying viral marketing opportunities. Existing methods often focus on limited cross-modal information within individual micro-videos, overlooking the potential advantages of exploiting vast repository of past videos. We present MMRA, a multi-modal retrieval-augmented popularity prediction model that enhances prediction accuracy using relevant retrieved information. MMRA first retrieves relevant instances from a multi-modal memory bank, aligning video and text through transformation mechanisms involving a vision model and a text-based retriever. Additionally, a multi-modal interaction network is carefully designed to jointly capture cross-modal correlations within the target video and extract informative knowledge through retrieved instances, ultimately enhancing the prediction. Extensive experiments conducted on the real-world micro-video dataset demonstrate the superiority of MMRA when compared to state-of-the-art models. The code and data are available at https://github.com/ICDM-UESTC/MMRA.

Information diffusion prediction, which aims to infer the infected behavior of individual users during information spread, is critical for understanding the dynamics of information propagation and users' influence on online social media. To date, existing methods either focus on capturing limited contextual information from a single cascade, overlooking the potentially complex dependencies across different cascades, or they are committed to improving model performance by using intricate technologies to extract additional features as supplements to user representations, neglecting the drift of model performance across different platforms. To address these limitations, we propose a novel framework called CARE (CAscade-REtrieved In-Context Learning) inspired by the concept of in-context learning in LLMs. Specifically, CARE first constructs a prompts pool derived from historical cascades, then utilizes ranking-based search engine techniques to retrieve prompts with similar patterns based on the query. Moreover, CARE also introduces two augmentation strategies alongside social relationship enhancement to enrich the input context. Finally, the transformed query-cascade representation from a GPT-type architecture is projected to obtain the prediction. Experiments on real-world datasets from various platforms show that CARE outperforms state-of-the-art baselines in terms of effectiveness and robustness in information diffusion prediction.

Landslides refer to occurrences of massive ground movements due to geological (and meteorological) factors, and can have disastrous impacts on property, economy, and even lead to the loss of life. The advances in remote sensing provide accurate and continuous terrain monitoring, enabling the study and analysis of land deformation which, in turn, can be used for land deformation prediction. Prior studies either rely on predefined factors and patterns or model static land observations without considering the subtle interactions between different point locations and the dynamic changes of the surface conditions, causing the prediction model to be less generalized and unable to capture the temporal deformation characteristics. To address these issues, we present DyLand, a dynamic manifold learning framework that models the dynamic structures of the terrain surface. We contribute to the land deformation prediction literature in four directions. First, DyLand learns the spatial connections of interferometric synthetic aperture radar (InSAR) measurements and estimates the conditional distributions on a dynamic terrain manifold with a novel normalizing flow-based method. Second, instead of modeling the stable terrains, we incorporate surface permutations and capture the innate dynamics of the land surface while allowing for tractable likelihood estimations on the manifold. Third, we formulate the spatiotemporal learning of land deformations as a dynamic system and unify the learning of spatial embeddings and surface deformation. Finally, extensive experiments on curated real-world InSAR datasets (land slopes prone to landslides) show that DyLand outperforms existing benchmark models.

Information cascade popularity prediction is an important problem in social network content diffusion analysis. Various facets have been investigated (e.g., diffusion structures and patterns, user influence) and, recently, deep learning models based on sequential architecture and graph neural network (GNN) have been leveraged. However, despite the improvements attained in predicting the future popularity, these methodologies fail to capture two essential aspects inherent to information diffusion: (1) the temporal irregularity of cascade event – i.e., users’ re-tweetings at random and non-periodic time instants; and (2) the inherent uncertainty of the information diffusion. To address these challenges, in this work, we present CasDO – a novel framework for information cascade popularity prediction with probabilistic diffusion models and neural ordinary differential equations (ODEs). We devise a temporal ODE network to generalize the discrete state transitions in RNNs to continuous-time dynamics. CasDO introduces a probabilistic diffusion model to consider the uncertainties in information diffusion by injecting noises in the forwarding process and reconstructing cascade embedding in the reversing process. Extensive experiments that we conducted on three large-scale datasets demonstrate the advantages of the CasDO model over baselines.

Fine-grained urban flow inference (FUFI) problem aims at inferring the high-resolution flow maps from the coarse-grained ones, which plays an important role in sustainable and economic urban computing and traffic management. Previous models addressed the FUFI problem from spatial constraint, external factors, and memory cost. However, utilizing the new urban flow maps to calibrate the learned model is very challenging due to the "catastrophic forgetting" problem and is still under-explored. In this paper, we make the first step in FUFI and present CUFAR -- Continual Urban Flow inference with Adaptive knowledge Replay -- a novel framework for inferring the fine-grained citywide traffic flows. Specifically, (1) we design a spatial-temporal inference network that can extract better flow map features from both local and global levels; (2) then we present an adaptive knowledge replay (AKR) training algorithm to selectively replay the learned knowledge to facilitate the learning process of the model on new knowledge without forgetting. In addition, we also propose a knowledge discriminator to avoid "negative replaying" issue introduced by noisy urban flow maps. Extensive experiments on four large-scale real-world FUFI datasets demonstrate that our proposed model consistently outperforms strong baselines and effectively mitigates the forgetting problem. Source code is available at: https://github.com/PattonYu/CUFAR.

Predicting information cascade popularity is a fundamental problem for understanding the nature of information propagation on social media. However, existing works fail to capture an essential aspect of information propagation: the temporal irregularity of cascade event -- i.e., users' re-tweetings at random and non-periodic time instants. In this work, we present a novel framework CasODE for information cascade prediction with neural ordinary differential equations (ODEs). CasODE generalizes the discrete state transitions in RNNs to continuous-time dynamics for modeling the irregular-sampled events in information cascades. Experimental evaluations on real-world datasets demonstrate the advantages of the CasODE over baseline approaches.

Earnings conference calls are indicative information events for volatility forecasting, which is essential for financial risk management and asset pricing. Although recent volatility forecasting models have explored the textual content of conference calls for prediction, they suffer from modeling the long-text and representing the risk-relevant information. This work proposes to identify key sentences for robust and interpretable transcript representation learning based on the cognitive theory. Specifically, we introduce TextRank to find key sentences and leverage attention mechanism to screen out the candidates by modeling the semantic correlations. Upon on the structural information of earning conference calls, we propose a structure-based contrastive learning method to facilitate the effective transcript representation. Empirical results on the benchmark dataset demonstrate the superiority of our model over competitive baselines in volatility forecasting.

Demographic biases and social stereotypes are common in pretrained language models (PLMs), while the fine-tuning in downstream applications can also produce new biases or amplify the impact of the original biases. Existing works separate the debiasing from the fine-tuning procedure, which results in a gap between intrinsic bias and application bias. In this work, we propose a debiasing framework CauDebias to eliminate both biases, which directly combines debiasing with fine-tuning and can be applied for any PLMs in downstream tasks. We distinguish the bias-relevant (non-causal factors) and label-relevant (causal factors) parts in sentences from a causal invariant perspective. Specifically, we perform intervention on non-causal factors in different demographic groups, and then devise an invariant risk minimization loss to trade-off performance between bias mitigation and task accuracy. Experimental results on three downstream tasks show that our CauDebias can remarkably reduce biases in PLMs while minimizing the impact on downstream tasks.

Source localization, as a reverse problem of graph diffusion, is important for many applications such as rumor tracking, detecting computer viruses, and finding epidemic spreaders. However, it is still under-explored due to the inherent uncertainty of the diffusion process: after a long period of propagation, the same diffusion process may start with diverse sources. Most existing solutions utilize deterministic models and therefore cannot describe the diffusion uncertainty of sources. Moreover, current probabilistic approaches are hard to conduct smooth transformations with variational inference. To overcome the limitations, we propose a probabilistic framework using continuous normalizing flows with invertible transformations and graph neural networks to explicitly model the uncertainty of the diffusion source. Experimental results on two real-world datasets demonstrate the effectiveness of our model over strong baselines.

Recent literature has shown that denoising diffusion probabilistic models (DDPMs) can be used to synthesize high-fidelity samples with a competitive (or sometimes better) quality than previous state-of-the-art approaches. However, few attempts have been made to apply DDPM for the speech enhancement task. The reported performance of the existing works is relatively poor and significantly inferior to other generative methods. In this work, we first reveal the difficulties in applying existing diffusion models to the field of speech enhancement. Then we introduce DR-DiffuSE, a simple and effective framework for speech enhancement using conditional diffusion models. We present three strategies (two in diffusion training and one in reverse sampling) to tackle the condition collapse and guarantee the sufficient use of condition information. For efficiency, we introduce the fast sampling technique to reduce the sampling process into several steps and exploit a refinement network to calibrate the defective speech. Our proposed method achieves the state-of-the-art performance to the GAN-based model and shows a significant improvement over existing DDPM-based algorithms.

The recent advance in graph neural networks (GNNs) has inspired a few studies to leverage the dependencies of variables for time series prediction. Despite the promising results, existing GNN-based models cannot capture the global dynamic relations between variables owing to the inherent limitation of their graph learning module. Besides, multi-scale temporal information is usually ignored or simply concatenated in prior methods, resulting in inaccurate predictions. To overcome these limitations, we present CGMF, a Continuous Graph learning method for Multivariate time series Forecasting (CGMF). Our CGMF consists of a continuous graph module incorporating differential equations to capture the long-range intra- and inter-relations of the temporal embedding sequence. We also introduce a controlled differential equation-based fusion mechanism that efficiently exploits multi-scale representations to form continuous evolutional dynamics and learn rich relations and patterns shared across different scales. Comprehensive experiments demonstrate the effectiveness of our method for a variety of datasets.

Learning domain-invariant representations is a major task of out-of-distribution generalization. To address this issue, recent efforts have taken into accounting causality, aiming at learning the causal factors with regard to tasks. However, extending existing generalization methods for adapting non-stationary time series may be ineffective, because they fail to model the underlying causal factors due to temporal-domain shifts except for source-domain shifts, as pointed out by recent studies. To this end, we propose a novel model DyCVAE to learn dynamic causal factors. The results on synthetic and real datasets demonstrate the effectiveness of our proposed model for the task of generalization in time series domain.

Demographic biases and social stereotypes are common in pretrained language models (PLMs), and a burgeoning body of literature focuses on removing the unwanted stereotypical associations from PLMs.However, when fine-tuning these bias-mitigated PLMs in downstream natural language processing (NLP) applications, such as sentiment classification, the unwanted stereotypical associations resurface or even get amplified.Since pretrain&fine-tune is a major paradigm in NLP applications, separating the debiasing procedure of PLMs from fine-tuning would eventually harm the actual downstream utility.In this paper, we propose a unified debiasing framework Causal-Debias to remove unwanted stereotypical associations in PLMs during fine-tuning.Specifically, Causal-Debias mitigates bias from a causal invariant perspective by leveraging the specific downstream task to identify bias-relevant and labelrelevant factors.We propose that bias-relevant factors are non-causal as they should have little impact on downstream tasks, while labelrelevant factors are causal.We perform interventions on non-causal factors in different demographic groups and design an invariant risk minimization loss to mitigate bias while maintaining task performance.Experimental results on three downstream tasks show that our proposed method can remarkably reduce unwanted stereotypical associations after PLMs are finetuned, while simultaneously minimizing the impact on PLMs and downstream applications. *Corresponding author unwanted stereotypical associations in PLMs.For example, some works (Zmigrod et al., 2019) pretrain a language model using original and counterfactual corpus in order to cancel-out biased associations, some works (Liang et al., 2020) focus on debiasing post-hoc sentence representations, and others (Guo et al., 2022;Cheng et al., 2021) design bias-equalizing objectives to fine-tune PLM's parameters.

Despite the growing use of the Somali language in various online domains, research on Somali language information retrieval remains limited and primarily relies on query translation due to the lack of a dedicated corpus.To address this problem, we collaborated with language experts and natural language processing (NLP) researchers to create an annotated corpus for Somali information retrieval.This corpus comprises 2335 documents collected from various well-known online sites, such as hiiraan online, dhacdo net, and Somali poetry books.We explain how the corpus was constructed, and develop a Somali language information retrieval system using a pseudo-relevance feedback (PRF) query expansion technique on the corpus.Note that collecting such a data set for the low-resourced Somali language can help overcome NLP barriers, such as the lack of electronically available data sets.Which, if available, can enable the development of various NLP tools and applications such as questionanswering and text classification.It also provides researchers with a valuable resource for investigating and developing new techniques and approaches for Somali.

The rising popularity of online social network services has attracted a lot of research focusing on mining various user patterns. Among them, accurate IP geolocation is essential for a plethora of location-aware applications. However, despite extensive research efforts and significant advances, the "accurate and reliable'' desideratum is yet to be achieved at a higher quality level. This work presents a graph neural network (GNN)-based model, called TrustGeo, for trustworthy street-level IP geolocation. A distinct and important aspect of TrustGeo is the incorporation of sources of uncertainty in the learning process. The results of our extensive experimental evaluations on three real-world datasets demonstrate the superiority of our framework in significantly improving the accuracy and trustworthiness of street-level IP geolocation. Our code and datasets are available at https://github.com/ICDM-UESTC/TrustGeo.

IP geolocation refers to the process of determining the geographic locations of Internet Protocol (IP) addresses, which is important for mobile computing and spatial data management. Despite extensive research efforts, a client-independent geolocation service with high accuracy and reliability has not yet been developed. This paper presents a graph neural network (GNN) model, dubbed RIPGeo, for robust street-level IP geolocation. Three factors that affect data quality are identified, and the importance of considering data quality in algorithm development is emphasized. Two novel self-supervised perturbational training strategies are proposed to enhance the generalization and robustness of the model. A multi-task learning framework is introduced to solve the homogenized representation problem caused by perturbational training, demonstrating much more efficiency than prevailing solutions. Theoretical analysis and experimental results demonstrate the superiority of our framework in significantly improving the accuracy and stability of street-level IP geolocation.

Speech enhancement (SE) aims to improve the intelligibility and quality of speech in the presence of non-stationary additive noise. Deterministic deep learning models have traditionally been used for SE, but recent studies have shown that generative approaches, such as denoising diffusion probabilistic models (DDPMs), can also be effective. However, incorporating condition information into DDPMs for SE remains a challenge. We propose a model-agnostic method called DOSE that employs two efficient condition-augmentation techniques to address this challenge, based on two key insights: (1) We force the model to prioritize the condition factor when generating samples by training it with dropout operation; (2) We inject the condition information into the sampling process by providing an informative adaptive prior. Experiments demonstrate that our approach yields substantial improvements in high-quality and stable speech generation, consistency with the condition factor, and inference efficiency. Codes are publicly available at https://github.com/ICDM-UESTC/DOSE.

Abstract The wide application of image classification has given rise to many intelligent systems, such as face recognition systems, which makes our life more convenient. However, the ensuing privacy leakage problem has become increasingly serious. The training of a deep neural network requires lots of data, which may contain sensitive information of users and may be exploited by data collectors. A perturbation algorithm named RRN is proposed for image data based on local differential privacy, which provides a rigorous privacy guarantee. Existing solutions have low accuracy due to the high sensitivity of an image; the authors' method combines the Randomized Response mechanism with the Laplace mechanism to solve this problem. Experiments were conducted on the MNIST and CIFAR‐10 datasets to show the effectiveness of the algorithm. Experimental results shows that the model is better than baseline models. The algorithm was also implemented on the commonly used model in deep learning, the VGG model, which can achieve 96.4% accuracy in the non‐private version on the CIFAR‐10 dataset. The accuracy of the differential private VGG model based on the RRN algorithm is 83% when , which is still excellent. The experimental results show that the RRN algorithm can both preserve privacy and data utility.

Fine-grained urban flow inference (FUFI) problem aims to infer the fine-grained flow maps from coarse-grained ones, benefiting various smart-city applications by reducing electricity, maintenance, and operation costs. Existing models use techniques from image super-resolution and achieve good performance in FUFI. However, they often rely on supervised learning with a large amount of training data, and often lack generalization capability and face overfitting. We present a new solution:Spatial-TemporalContrasting for Fine-Grained UrbanFlow Inference (STCF). It consists of (i) two pre-training networks for spatial-temporal contrasting between flow maps; and (ii) one coupled fine-tuning network for fusing learned features. By attractingspatial-temporally similarflow maps while distancing dissimilar ones within the representation space, STCF enhances efficiency and performance. Comprehensive experiments on two large-scale, real-world urban flow datasets reveal that STCF reduces inference error by up to 13.5%, requiring significantly fewer data and model parameters than prior arts.

Many deep (semi-) supervised neural network-based methods have been proposed for anomaly detection, tackling the issue of limited labeled data. They have shown good performance but still face two major challenges. First, insufficient labeled data limits their flexibility. Second, measuring the uncertainty of the prediction, especially when dealing with objects deviating largely from training data, has not been well studied. Another common reason preventing them from prevailing is that they learn a determined function to make predictions from the input. This usually makes the predicted results uncertain and lacks robustness. To address these problems, we propose a novel framework, incorporating the neural process into the semi-supervised anomaly detection paradigm and efficiently using unlabeled data and a handful of labeled data in training. Different from other methods, ours is equivalent to modeling the distribution of functions representing anomalous patterns according to the labeled data rather than learning a single determined function for anomaly detection. Our approach improves the flexibility and robustness under the condition of insufficient training data, and can measure the uncertainty of prediction results. Extensive experiments under real-world datasets demonstrate that our proposed method can significantly improve anomaly detection performance compared to several cutting-edge benchmarks.

Identifying the geolocation of social media users is an important problem in a wide range of applications, spanning from disease outbreaks, emergency detection, local event recommendation, to fake news localization, online marketing planning, and even crime control and prevention. Researchers have attempted to propose various models by combining different sources of information, including text, social relation, and contextual data, which indeed has achieved promising results. However, existing approaches still suffer from certain constraints, such as: 1) a very few samples are available and 2) prediction models are not easy to be generalized for users from new regions-which are challenges that motivate our study. In this article, we propose a general framework for identifying user geolocation-MetaGeo, which is a meta-learning-based approach, learning the prior distribution of the geolocation task in order to quickly adapt the prediction toward users from new locations. Different from typical meta-learning settings that only learn a new concept from few-shot samples, MetaGeo improves the geolocation prediction with conventional settings by ensembling numerous mini-tasks. In addition, MetaGeo incorporates probabilistic inference to alleviate two issues inherent in training with few samples: location uncertainty and task ambiguity. To demonstrate the effectiveness of MetaGeo, we conduct extensive experimental evaluations on three real-world datasets and compare the performance with several state-of-the-art benchmark models. The results demonstrate the superiority of MetaGeo in both the settings where the predicted locations/regions are known or have not been seen during training.

Landslides refer to occurrences of massive ground movements due to geological (and meteorological) factors, and can have disastrous impact on property, economy, and even lead to loss of life. The advances of remote sensing provide accurate and continuous terrain monitoring, enabling the study and analysis of land deformation which, in turn, can be used for possible landslides forecast. Prior studies either rely on independent observations for displacement prediction or model static land characteristics without considering the subtle interactions between different locations and the dynamic changes of the surface conditions. We present DyLand -- Dynamic Manifold Learning with Normalizing Flows for Land deformation prediction -- a novel framework for learning dynamic structures of terrain surface and improving the performance of land deformation prediction. DyLand models the spatial connections of InSAR measurements and estimates conditional distributions of deformations on the terrain manifold with a novel normalizing flow-based method. Instead of modeling the stable terrains, it incorporates surface permutations and captures the innate dynamics of the land surface while allowing for tractable likelihood estimates on the manifold. Our extensive evaluations on curated InSAR datasets from continuous monitoring of slopes prone to landslides show that DyLand outperforms existing bechmarking models.

This paper exploits self-supervised learning (SSL) to learn more accurate and robust representations from the user-item interaction graph. Particularly, we propose a novel SSL model that effectively leverages contrastive multi-view learning and pseudo-siamese network to construct a pre-training and post-training framework. Moreover, we present three graph augmentation techniques during the pre-training stage and explore the effects of combining different augmentations, which allow us to learn general and robust representations for the GNN-based recommendation. Simple experimental evaluations on real-world datasets show that the proposed solution significantly improves the recommendation accuracy, especially for sparse data, and is also noise resistant.

The development of InSAR (satellite Interferometric Synthetic Aperture Radar) enables accurate monitoring of land surface deformations, and has led to advances of deformation forecast for preventing landslide, which is one of the severe geological disasters. Despite the unparalleled success, existing spatio-temporal models typically make predictions on static adjacency relationships, simplifying the conditional dependencies and neglecting the distributions of variables. To overcome those limitations, we propose a Distribution Aware Probabilistic Framework (DAPF), which learns manifold embeddings while maintaining the distribution of deformations. We obtain a dynamic adjacency matrix upon which we approximate the true posterior while emphasizing the spatio-temporal characteristics. Experimental results on real-world dataset validate the superior performance of our method.

Electricity forecasting has important implications for the key decisions in modern electricity systems, ranging from power generation, transmission, distribution and so on. In the literature, traditional statistic approaches, machine-learning methods and deep learning (e.g., recurrent neural network) based models are utilized to model the trends and patterns in electricity time-series data. However, they are restricted either by their deterministic forms or by independence in probabilistic assumptions -- thereby neglecting the uncertainty or significant correlations between distributions of electricity data. Ignoring these, in turn, may yield error accumulation, especially when relying on historical data and aiming at multi-step prediction. To overcome these, we propose a novel method named Probabilistic Electricity Forecasting (PrEF) by proposing a non-linear neural state space model (SSM) and incorporating copula-augmented mechanism into that, which can learn uncertainty-dependencies knowledge and understand interactive relationships between various factors from large-scale electricity time-series data. Our method distinguishes itself from existing models by its traceable inference procedure and its capability of providing high-quality probabilistic distribution predictions. Extensive experiments on two real-world electricity datasets demonstrate that our method consistently outperforms the alternatives.

Conditional neural process (CNP) has been extensively applied into data analyzing tasks due to its excellent ability to make accurate predictions for incomplete data points. However, in literature there are only few works that studied the CNPin recommendation systems. In this work, we propose CCFP, which is a collaborative filtering method that differs from other CF models by incorporating CNP into encoder-decoder architecture. By analyzing the complete user-item interaction data, our model fits a global representation that can better rep-resenting the features of users and items. CCFP can significantly improve the recommendation performance compared to baselines by predicting items for the target users with their incomplete observation data.

Fine-grained urban flow inference (FUFI) aims to infer the coarse-grained (CG) urban flow map to the corresponding fine-grained (FG) one, which plays an important role in efficient traffic monitoring and management in smart cities. In FUFI, the CG map can be obtained with only a small number of monitoring devices, greatly reducing the overhead of deploying devices and the costs of maintenance, labor, and electricity. Existing FUFI methods are mainly based on techniques from image super-resolution (SR) models, which cannot fully consider the influence of external factors and face the ill-posed problem in SR tasks. In this paper, we propose UFI-Flow, a novel approach for addressing the FUFI problem by learning the conditional distributions of CG and FG map pairs. Given the CG map and the latent variables, the corresponding FG map is inferred by invertible transformations. In addition, an augmented distribution fusion mechanism is further proposed to constrain the urban flow distribution within the influence of external factors. We provide a new large-scale real-world FUFI dataset and show that UFI-Flow significantly outperforms the strong baselines.

Information cascade is typically formalized as a process of (simplified) discrete sequence of events, and recent approaches have tackled its prediction via variants of recurrent neural networks. However, the information diffusion process is essentially an evolving directed acyclic graph (DAG) in the continuous-time domain. In this paper, we propose a transformer enhanced Hawkes process (Hawkesformer), which links the hierarchical attention mechanism with Hawkes process to model the arrival stream of discrete events continuously. A two-level attention architecture is used to parameterize the intensity function of Hawkesformer, which captures the long-term dependencies between nodes in graph and better embeds the cascade evolution rate for modeling short-term outbreaks. Experimental results demonstrate the significant improvements of Hawkesformer over the state-of-the-art.

Fine-grained urban flow inference (FUFI) aims at enhancing the resolution of traffic flow, which plays an important role in intelligent traffic management. Existing FUFI methods are mainly based on techniques from image super-resolution (SR) models, which cannot fully capture the influence of external factors and face the ill-posed problem in SR tasks. In this paper, we propose UFI-Flow – Urban Flow Inference via normalizing Flow, a novel model for addressing the FUFI problem in a principled manner by using a single probabilistic loss. UFI-Flow therefore directly accounts for the ill-posed nature of the problem and learns spatial correlations between urban flow maps. In addition, an augmented distribution fusion mechanism is further proposed to reinforce the influence of external factors in the joint distribution inference. We conduct comprehensive experiments on real-world datasets to show the superiority of the proposed model compared to the state-of-the-art baseline approaches.

Graph Neural Networks (GNNs) and its many variants have recently been successfully utilized to tackle various recommendation tasks. Although effective, existing methods still face several limitations. First, supervision signal sparsity makes it difficult for them to learn high-quality representations and optimize the model parameters. Second, the learned representations are vulnerable to noisy interactions, as the neighborhood aggregation scheme increases the impact of observed edges. To alleviate these issues, we exploit self-supervised learning (SSL) to learn more accurate and robust representations from the user-item interaction graph. Particularly, we propose a novel SSL model that effectively integrates contrastive multi-view learning and pseudo-siamese network to construct a pre-training and post-training framework. Moreover, we present three data augmentation techniques during the pre-training stage and explore the effects of combining different augmentations, which allow us to learn more general and robust representations for the recommendation. Experimental evaluations on two real-world datasets show that the proposed solution significantly improves the recommendation accuracy, especially for sparse supervision signal, and is also noise resistant.

A large amount of disinformation on social media has penetrated into various domains and brought significant adverse effects. Understanding their roots and propagation becomes desired in both academia and industry. Prior literature has developed many algorithms to identify this disinformation, particularly rumor detection. Some leverage the power of deep learning and have achieved promising results. However, they all focused on building predictive models and improving forecast accuracy, while two important factors - popularity and conformity biases - that play critical roles in rumor spreading behaviors are usually neglected.To overcome such an issue and alleviate the bias from these two factors, we propose a rumor detection framework to learn debiased user preference and effective event representation in a causal view. We first build a graph to capture causal relationships among users, events, and their interactions. Then we apply the causal intervention to eliminate popularity and conformity biases and obtain debiased user preference representation. Finally, we leverage the power of graph neural networks to aggregate learned user representation and event features for the final event type classification. Empirical experiments conducted on two real-world datasets demonstrate the effectiveness of our proposed approach compared to several cutting-edge baselines.

Modeling complex spatial and temporal dependencies are indispensable for location-bound time series learning. Existing methods, typically relying on graph neural networks (GNNs) and temporal learning modules based on recurrent neural networks, have achieved significant performance improvements. However, their representation capabilities and prediction results are limited when pre-defined graphs are unavailable. Unlike spatio-temporal GNNs focusing on designing complex architectures, we propose a novel adaptive graph construction strategy: Self-Paced Graph Contrastive Learning (SPGCL). It learns informative relations by maximizing the distinguishing margin between positive and negative neighbors and generates an optimal graph with a self-paced strategy. Specifically, the existing neighborhoods iteratively absorb more reliable nodes with the highest affinity scores as new neighbors to generate the next-round neighborhoods, and augmentations are applied to improve the transferability and robustness. As the adaptively self-paced graph approaches the optimized graph for prediction, the mutual information between nodes and the corresponding neighbors is maximized. Our work provides a new perspective of addressing spatio-temporal learning problems beyond information aggregation in Euclidean space and can be generalized to different tasks. Extensive experiments conducted on two typical spatio-temporal learning tasks (traffic forecasting and land displacement prediction) demonstrate the superior performance of SPGCL against the state-of-the-art.

Forecasting complex time series is ubiquitous and vital in a range of applications but challenging. Recent advances endeavor to achieve progress by incorporating various deep learning techniques (e.g., RNN and Transformer) into sequential models. However, clear patterns are still hard to extract since time series are often composed of several intricately entangled components. Motivated by the success of disentangled variational autoencoder in computer vision and classical time series decomposition, we plan to infer a couple of representations that depict seasonal and trend components of time series. To achieve this goal, we propose LaST, which, based on variational inference, aims to disentangle the seasonal-trend representations in the latent space. Furthermore, LaST supervises and disassociates representations from the perspectives of themselves and input reconstruction, and introduces a series of auxiliary objectives. Extensive experiments prove that LaST achieves state-of-the-art performance on time series forecasting task against the most advanced representation learning and end-to-end forecasting models. For reproducibility, our implementation is publicly available on Github.

Most existing studies that employ social-trust information to solve the data sparsity issue in recommender systems assume that socially connected users have equal influence on each other. However, this assumption does not hold in practice since users and their friends may not have similar interests because social connections are multifaceted and exhibit heterogeneous strengths in different scenarios. Therefore, estimating the diverse levels of influence among entities (users/items/social connections) is very important in advancing social recommender systems. Towards this goal, we propose a new model named Social-Trust-Aware Variational Recommendation (SOAP-VAE). Particularly, SOAP-VAE leverages graph attention network techniques to capture the varying levels of influence and the complex interaction patterns among all the entities collectively and holistically. In doing so, heterogeneity among entities is obtained seamlessly. Consequently, we generate social-trust-aware item embedding representations in which the right level of influence has been integrated. Next, based on these rich social-trust-aware item representations, we formulate the first-ever social-trust-aware prior in literature. Unlike priors utilized in earlier VAE-based recommendation models, this novel prior aids in dealing with the issue of posterior-collapse and can effectively capture the uncertainty of latent space. In effect, the model produces better latent representations, which significantly alleviates the data sparsity issue. Finally, we empirically show that SOAP-VAE outperforms several state-of-the-art baselines on three real-world data sets.

We present HydroFlow, a novel deep generative model for predicting the electricity generation demand of large-scale hydropower stations. HydroFlow uses a latent stochastic recurrent neural network to capture the dependencies in the multivariate time series. It not only utilizes the hidden state of the neural network, but also considers the uncertainty of variables related to natural and social factors. We also introduce an end-to-end approach based on generative flows to approximate the posterior distribution of time series with exact likelihoods. Our model is powerful as adding stochasticity to different factors (e.g., reservoir capacity and water-flow measurements) and thus overcomes the expressiveness limitations of deterministic prediction methods. It also enables trainable latent transformations that can improve the model interpretability. We evaluate HydroFlow on the data collected from the hydropower stations of a large-scale hydropower development company. Experimental results show that our model significantly outperforms the state-of-the-art baseline methods while providing explainable results.

Stockmovement prediction using Tweets (text) and historical price signals (time series) remains a challenging task due to the complex, noisy and dynamic nature of the stock market. The key to improve prediction performance is effectively capturing the complementarity between market sentiment signal from Tweets and time-series signal from the stock price. In this paper, we contribute a new solutionStockNFby exploiting a deep generative model technique, Normalizing Flow (NF), to learn more flexible and expressive posterior distributions of latent variables of Tweets and price signals, which can largely ameliorate the bias inference problem in existing methods. We empirically evaluate the NF technique on a public stock movement prediction dataset and show that StockNF outperforms the state-of-the-art baselines.

Despite offering efficient solutions to a plethora of novel challenges, existing approaches on mobility modeling require a large amount of labeled data when training effective and application-specific models. This renders them inapplicable to certain scenarios, where only a few samples are observed, and data types are unseen during training. To address these issues, we present a novel mobility learning method-MetaMove, the first metalearning-based model generalizing mobility prediction and classification in a unified framework. MetaMove deals with the problem of training for unseen mobility patterns by generalizing from the known patterns. It trains the model over a variety of patterns sampled from different users and optimizes it on their distribution. To update and fine tune the individual pattern learners, we employ a fast adapting model-agnostic method for very few available trajectory samples. MetaMove exploits unlabeled trajectory data at both metatraining and adaptation levels, thereby alleviating the problem of data sparsity while enforcing less sensitivity to negative samples. We conducted extensive experiments to demonstrate its effectiveness and efficiency on two practical applications-motion trace discrimination and next check-in prediction. The results demonstrated significant improvements of MetaMove over the state-of-the-art benchmarks.

In theory, the variational auto-encoder (VAE) is not suitable for recommendation tasks, although it has been successfully utilized for collaborative filtering (CF) models. In this paper, we propose a Gaussian Copula-Vector Quantized Autoencoder (GC-VQAE) model that differs prior arts in two key ways: (1) Gaussian Copula helps to model the dependencies among latent variables which are used to construct a more complex distribution compared with the mean-field theory; and (2) by incorporating a vector quantisation method into encoders our model can learn discrete representations which are consistent with the observed data rather than directly sampling from the simple Gaussian distributions. Our approach is able to circumvent the "posterior collapse'' issue and break the prior constraint to improve the flexibility of latent vector encoding and learning ability. Empirically, GC-VQAE can significantly improve the recommendation performance compared to existing state-of-the-art methods.

Monitoring urban flow timely and accurately is crucial for many industrial applications – from urban planning to traffic control in the smart cities. This work introduces a new method for inferring fine-grained urban flow with the internet of mobile things such as taxis and bikes. We tackle the problem from a new perspective and present a novel deep learning method UrbanODE (Urban flow inference with Neural Ordinary Differential Equations). Furthermore, UrbanODE provides a flexible balance between flow inference accuracy and computational efficiency, which is important in computation restricted scenarios such as pervasive edge computing. Extensive evaluations on real-world traffic flow data demonstrate the superiority of the proposed method.

Traffic forecasting is of particular interest in intelligent transportation systems (ITS). This problem is challenging owing to the complicated spatio-temporal dependencies between different areas in a road sensor network. Previous approaches have applied various deep learning methods for traffic forecasting, e.g., leveraging graph convolutional networks (GCNs) for spatial correlation modeling and utilizing recurrent neural networks (RNNs) to capture temporal traffic evolutions. However, the existing GCN-based models can not adequately distinguish the non-Euclidean topological structure of road traffic and are easily affected by random traffic noise. This work proposes an end-to-end framework to capture spatial dependencies through graph isomorphism network, while explicitly taking network topologic similarities into account and leveraging symmetric traffic for learning the traffic conditions. Extensive experiments on two real-world traffic datasets demonstrate the superiority of our proposed approach.

Intelligent traffic signal control is crucial for efficient transportation systems. Recent studies use reinforcement learning (RL) to coordinate traffic signals and improve traffic signal cooperation. However, they either design the state of agents in a heuristic manner or model traffic dynamics in a deterministic way. This work presents a variational graph learning model TSC-GNN (Traffic Signal Control via probabilistic Graph Neural Networks) to learn the latent representations of agents and generate Q-value while taking traffic uncertainty conditions into account. Besides, we explain the rationality behind our state design using transportation theory. Experimental results conducted on real-world datasets demonstrate our model’s superiority, e.g., it achieves more than 8% traffic efficiency improvement compared with the state-of-the-art baselines.

Network alignment (NA) aims to link common nodes across multiple networks and is an essential task in many graph mining applications. Despite the progress achieved by many recent works, several fundamental limitations have eluded the proper cohesive way of addressing, including matching confusion, lack of the formal treatment of uncertainty, and Point-to-Point (P2P) constraint. This study proposes a novel framework UANA (Uncertainty-Aware Network Alignment) to tackle the limitations of the existing works. By embedding nodes as Gaussian distributions rather than point vectors, UANA enables to capture the uncertainty of a node representation, while being able to discriminate the anchor nodes from the potentially confusing neighbors. We address the P2P matching constraint by introducing an adversarial learning paradigm, which relaxes the exact matching assumption during training with an across-domain generative procedure to reduce the matching errors on testing nodes. In the end, interpretability methods are included to explain the aligning results made by our UANA based on the robust statistics, which enables the explanation of the effect of individual training sample on the NA performance without the need of retraining the model. Extensive experiments conducted on real-world data sets demonstrate that UANA significantly outperforms existing state-of-the-art baselines while providing explainable results.

As one of the most important applications of industrial Internet of Things, intelligent transportation system aims to improve the efficiency and safety of transportation networks. In this article, we propose a novel Bayesian framework entitled variational graph recurrent attention neural networks (VGRAN) for robust traffic forecasting. It captures time-varying road-sensor readings through dynamic graph convolution operations and is capable of learning latent variables regarding the sensor representation and traffic sequences. The proposed probabilistic method is a more flexible generative model considering the stochasticity of sensor attributes and temporal traffic correlations. Moreover, it enables efficient variational inference and faithful modeling of implicit posteriors of traffic data, which are usually irregular, spatial correlated, and multiple temporal dependents. Extensive experiments conducted on two real-world traffic datasets demonstrate that the proposed VGRAN model outperforms state-of-the-art approaches while capturing innate ambiguity of the predicted results.

Identifying user geolocation in online social networks is an essential task in many locationbased applications.Existing methods rely on the similarity of text and network structure, however, they suffer from a lack of interpretability on the corresponding results, which is crucial for understanding model behavior.In this work, we adopt influence functions to interpret the behavior of GNN-based models by identifying the importance of training users when predicting the locations of the testing users.This methodology helps with providing meaningful explanations on prediction results.Furthermore, it also initiates an attempt to uncover the so-called "black-box" GNN-based models by investigating the effect of individual nodes.

Modeling the information diffusion process is an essential step towards understanding the mechanisms driving the success of information. Existing methods either exploit various features associated with cascades to study the underlying factors governing information propagation, or leverage graph representation techniques to model the diffusion process in an end-to-end manner. Current solutions are only valid for a static and fixed observation scenario and fail to handle increasing observations due to the challenge of catastrophic forgetting problems inherent in the machine learning approaches used for modeling and predicting cascades. To remedy this issue, we propose a novel dynamic information diffusion model CICP (Continual Information Cascades Prediction). CICP employs graph neural networks for modeling information diffusion and continually adapts to increasing observations. It is capable of capturing the correlations between successive observations while preserving the important parameters regarding cascade evolution and transition. Experiments conducted on real-world cascade datasets demonstrate that our method not only improves the prediction performance with accumulated data but also prevents the model from forgetting previously trained tasks.

Predicting the potential user interest on topics in online social networks is important for many practical applications such as advertising, recommendation and malicious account identification. Previous methods on such topic prediction problem mainly focus on learning user preference from historical posting content, and/or rely on the interest of friends to infer the topics a user may be interested in. However, these methods fail to take full advantage of high-order interactions between users and topics and the implicit relations among users, which may result in limited performance. In addition, existing approaches usually require a large amount of samples to train the model and therefore have poor prediction performance for the users who have few content and/or rarely follow the topics. To overcome these limitations, we present a novel method MetaTP (Meta learning based Topic Prediction) for exploiting the complex preference of users over the topics and identify the potential topics for cold-start users. MetaTP is built on a fast graph convolutional network to estimate the user interest through extracting user posting behavior from historical posting content and recursively aggregating the interest from the social friends of a user. Moreover, MetaTP introduces a new way of training prediction model in a meta-learning manner, which not only improves the performance on topic prediction but also can effectively and efficiently adapt to users with a few records. We validate our MetaTP model on real-world datasets crawled from popular social platforms and the empirical results show that our approach significantly outperforms the state-of-the-art baselines.

User identity linkage (UIL) aims to link identical users engaging in multiple social networks. It has received considerable attention in both academia and industry due to its profound implications for multiple applications. Although existing approaches have achieved promising progress in UIL using various graph learning methods, they usually require a large number of labeled anchor nodes which, however, are difficult to obtain in real-world social platforms due to privacy issues. We introduce a novel UIL model NWUIL (Network Wasserstein learning for UIL) to identify anchor users across social networks in a fully unsupervised manner. Instead of point vector embedding of nodes as in previous methods, NWUIL captures node distribution in Wasserstein space with graph neural networks. We also propose to reformulate the UIL task as an optimal network transport problem, and then introduce an unsupervised mapping process based on the network Wasserstein distance for UIL. In this way, our method not only improves the anchor node aligning accuracy but also alleviates the issues caused by insufficient labeled anchor nodes. We conduct extensive experiments using real-world datasets, and demonstrate that NWUIL significantly outperforms existing unsupervised baselines while showing competitive performance as some state-of-the-art supervised approaches.

Flow super-resolution (FSR) enables inferring fine-grained urban flows with coarse-grained observations and plays an important role in traffic monitoring and prediction. The existing FSR solutions rely on deep CNN models (e.g., ResNet) for learning spatial correlation, incurring excessive memory cost and numerous parameter updates. We propose to tackle the urban flows inference using dynamic systems paradigm and present a new method FODE -- FSR with Ordinary Differential Equations (ODEs). FODE extends neural ODEs by introducing an affine coupling layer to overcome the problem of numerically unstable gradient computation, which allows more accurate and efficient spatial correlation estimation, without extra memory cost. In addition, FODE provides a flexible balance between flow inference accuracy and computational efficiency. A FODE-based augmented normalization mechanism is further introduced to constrain the flow distribution with the influence of external factors. Experimental evaluations on two real-world datasets demonstrate that FODE significantly outperforms several baseline approaches.

Network alignment (NA) - i.e., linking entities from different networks (also known as identity linkage) - is a fundamental problem in many application domains. Recent advances in deep graph learning have inspired various auspicious approaches for tackling the NA problem. However, most of the existing works suffer from efficiency and generalization, due to complexities and redundant computations.We approach the NA from a different perspective, tackling it via meta-learning in a semi-supervised manner, and propose an effective and efficient approach called Meta-NA - a novel, conceptually simple, flexible, and general framework. Specifically, we reformulate NA as a one-shot classification problem and address it with a graph meta-learning framework. Meta-NA exploits the meta-metric learning from known anchor nodes to obtain latent priors for linking unknown anchor nodes. It contains multiple sub-networks corresponding to multiple graphs, learning a unified metric space, where one can easily link entities across different graphs. In addition to the performance lift, Meta-NA greatly improves the anchor linking generalization, significantly reduces the computational overheads, and is easily extendable to multi-network alignment scenarios. Extensive experiments conducted on three real-world datasets demonstrate the superiority of Meta-NA over several state-of-the-art baselines in terms of both alignment accuracy and learning efficiency.

Understanding in-network information diffusion is a fundamental problem in many application domains and one of the primary challenges is to predict the size of the information cascade. Most of the existing models rely either on hypothesized point process (e.g., Poisson and Hawkes process), or simply predict the information propagation via deep neural networks. However, they fail to simultaneously capture the underlying structure of a cascade graph and the propagation of uncertainty in the diffusion, which may result in unsatisfactory prediction performance. To address these, in this work we propose a novel probabilistic cascade prediction framework: Variational Cascade (VaCas) graph learning networks. VaCas allows a non-linear information diffusion inference and models the information diffusion process by learning the latent representation of both the structural and temporal information. It is a pattern-agnostic model leveraging variational inference to learn the node-level and cascade-level latent factors in an unsupervised manner. In addition, VaCas is capable of capturing both the cascade representation uncertainty and node infection uncertainty, while enabling hierarchical pattern learning of information diffusion. Extensive experiments conducted on real-world datasets demonstrate that VaCas significantly improves the prediction accuracy, compared to state-of-the-art approaches, while also enabling interpretability.

Hydropower is the largest renewable energy source for electricity generation in the world, with numerous benefits in terms of: environment protection (near-zero air pollution and climate impact), cost-effectiveness (long-term use, without significant impacts of market fluctuation), and reliability (quickly respond to surge in demand). However, the effectiveness of hydropower plants is affected by multiple factors such as reservoir capacity, rainfall, temperature and fluctuating electricity demand, and particularly their complicated relationships, which make the prediction/recommendation of station operational output a difficult challenge. In this paper, we present DeepHydro, a novel stochastic method for modeling multivariate time series (e.g., water inflow/outflow and temperature) and forecasting power generation of hydropower stations. DeepHydro captures temporal dependencies in co-evolving time series with a new conditioned latent recurrent neural networks, which not only considers the hidden states of observations but also preserves the uncertainty of latent variables. We introduce a generative network parameterized on a continuous normalizing flow to approximate the complex posterior distribution of multivariate time series data, and further use neural ordinary differential equations to estimate the continuous-time dynamics of the latent variables constituting the observable data. This allows our model to deal with the discrete observations in the context of continuous dynamic systems, while being robust to the noise. We conduct extensive experiments on real-world datasets from a large power generation company consisting of cascade hydropower stations. The experimental results demonstrate that the proposed method can effectively predict the power production and significantly outperform the possible candidate baseline approaches.

The advances in the Internet of Things (IoT) and increased availability of the road sensors allow for fine-grained traffic forecasting, which is of particular importance toward building an intelligent transportation system. In the literature, recent efforts have applied various deep learning methods for traffic forecasting, e.g., leveraging graph convolutional networks (GCNs) for spatial dependency modeling, and utilizing recurrent neural networks (RNNs) for capturing temporal dynamics. However, most of the existing approaches assume that spatial correlations are static and temporal correlations have only sequential dependencies and do not consider temporal periodicity of traffic across multiple time steps. The real challenge lies in using the dynamic spatiotemporal correlations while also considering the influence of the nontraffic-related factors, such as time-of-day and weekday-or-weekend in the learning architectures. We propose a novel framework titled “reinforced spatial-temporal attention graph (RSTAG) neural networks” for traffic prediction. Our method captures dynamic spatial correlations through diffusion network graphs, while temporal dependencies are represented through the sequence-to-sequence model with an attention mechanism. In addition, we utilize the policy gradient to update the model parameters while largely alleviating the exposure bias issue that exists in previous traffic prediction models. We conduct extensive experiments on two large-scale traffic data sets collected from the road sensor networks in Los Angles and Bay Area of California. The results demonstrate that our method significantly outperforms the state-of-the-art baselines.

Meta-learning has received a tremendous recent attention as a possible approach for mimicking human intelligence, i.e., acquiring new knowledge and skills with little or even no demonstration. Most of the existing meta-learning methods are proposed to tackle few-shot learning problems such as image and text, in rather Euclidean domain. However, there are very few works applying meta-learning to non-Euclidean domains, and the recently proposed graph neural networks (GNNs) models do not perform effectively on graph few-shot learning problems. Towards this, we propose a novel graph meta-learning framework -- Meta-GNN -- to tackle the few-shot node classification problem in graph meta-learning settings. It obtains the prior knowledge of classifiers by training on many similar few-shot learning tasks and then classifies the nodes from new classes with only few labeled samples. Additionally, Meta-GNN is a general model that can be straightforwardly incorporated into any existing state-of-the-art GNN. Our experiments conducted on three benchmark datasets demonstrate that our proposed approach not only improves the node classification performance by a large margin on few-shot learning problems in meta-learning paradigm, but also learns a more general and flexible model for task adaption.

In this work we propose DeepTrip -- an end-to-end method for better understanding of the underlying human mobility and improved modeling of the POIs' transitional distribution in human moving patterns. DeepTrip consists of: a Trip Encoder to embed a given route into a latent variable with a recurrent neural network (RNN); and a Trip Decoder to reconstruct this route conditioned on an optimized latent space. Simultaneously, we define an Adversarial Net composed of a generator and critic, which generates a representation for a given query and uses a critic to distinguish the trip representation generated from Trip Encoder and query representation obtained from Adversarial Net. DeepTrip enables regularizing the latent space and generalizing users' complex check-in preference. We demonstrate the effectiveness and efficiency of the proposed model, and the experimental evaluations show that DeepTrip outperforms the state-of-the-art baselines on various evaluation metrics.

Effectively predicting the size of an information cascade is critical for many applications spanning from identifying viral marketing and fake news to precise recommendation and online advertising. Traditional approaches either heavily depend on underlying diffusion models and are not optimized for popularity prediction, or use complicated hand-crafted features that cannot be easily generalized to different types of cascades. Recent generative approaches allow for understanding the spreading mechanisms, but with unsatisfactory prediction accuracy. To capture both the underlying structures governing the spread of information and inherent dependencies between re-tweeting behaviors of users, we propose a semi-supervised method, called Recurrent Cascades Convolutional Networks (CasCN), which explicitly models and predicts cascades through learning the latent representation of both structural and temporal information, without involving any other features. In contrast to the existing single, undirected and stationary Graph Convolutional Networks (GCNs), CasCN is a novel multi-directional/dynamic GCN. Our experiments conducted on real-world datasets show that CasCN significantly improves the prediction accuracy and reduces the computational cost compared to state-of-the-art approaches.

Network alignment (NA) is a fundamental problem in many application domains - from social networks, through biology and communications, to neuroscience. The main objective is to identify common nodes and most similar connections across multiple networks (resp. graphs). Many of the existing efforts focus on efficient anchor node linkage by leveraging various features and optimizing network mapping functions with the pairwise similarity between anchor nodes. Despite the recent advances, there still exist two kinds of challenges: (1) entangled node embeddings, arising from the contradictory goals of NA: embedding proximal nodes in a closed form for representation in a single network vs. discriminating among them when mapping the nodes across networks; and (2) lack of interpretability about the node matching and alignment, essential for understanding prediction tasks. We propose d NAME (disentangled Network Alignment with Matching Explainability) - a novel solution for NA in heterogeneous networks settings, based on a matching technique that embeds nodes in a disentangled and faithful manner. The NA task is cast as an adversarial optimization problem which learns a proximity-preserving model locally around the anchor nodes, while still being discriminative. We also introduce a method to explain our semi-supervised model with the theory of robust statistics, by tracing the importance of each anchor node and its explanations on the NA performance. This is extensible to many other NA methods, as it provides model interpretability. Experiments conducted on several public datasets show that d NAME outperforms the state-of-the-art methods in terms of both network alignment precision and node matching ranking.

Effectively modeling and predicting the information cascades is at the core of understanding the information diffusion, which is essential for many related downstream applications, such as fake news detection and viral marketing identification. Conventional methods for cascade prediction heavily depend on the hypothesis of diffusion models and hand-crafted features. Owing to the significant recent successes of deep learning in multiple domains, attempts have been made to predict cascades by developing neural networks based approaches. However, the existing models are not capable of capturing both the underlying structure of a cascade graph and the node sequence in the diffusion process which, in turn, results in unsatisfactory prediction performance. In this paper, we propose a deep multi-task learning framework with a novel design of shared-representation layer to aid in explicitly understanding and predicting the cascades. As it turns out, the learned latent representation from the shared-representation layer can encode the structure and the node sequence of the cascade very well. Our experiments conducted on real-world datasets demonstrate that our method can significantly improve the prediction accuracy and reduce the computational cost compared to state-of-the-art baselines.

We present a novel generative Session-Based Recommendation (SBR) framework, called VAriational SEssion-based Recommendation (VASER) - a non-linear probabilistic methodology allowing Bayesian inference for flexible parameter estimation of sequential recommendations. Instead of directly applying extended Variational AutoEncoders (VAE) to SBR, the proposed method introduces normalizing flows to estimate the probabilistic posterior, which is more effective than the agnostic presumed prior approximation used in existing deep generative recommendation approaches. VASER explores soft attention mechanism to upweight the important clicks in a session. We empirically demonstrate that the proposed model significantly outperforms several state-of-the-art baselines, including the recently-proposed RNN/VAE-based approaches on real-world datasets.

Unveiling human mobility patterns is an important task for many downstream applications like point-of-interest (POI) recommendation and personalized trip planning. Compelling results exist in various sequential modeling methods and representation techniques. However, discovering and exploiting the context of trajectories in terms of abstract topics associated with the motion can provide a more comprehensive understanding of the dynamics of patterns. We propose a new paradigm for moving pattern mining based on learning trajectory context, and a method - Context-Aware Variational Trajectory Encoding and Human Mobility Inference (CATHI) - for learning user trajectory representation via a framework consisting of: (1) a variational encoder and a recurrent encoder; (2) a variational attention layer; (3) two decoders. We simultaneously tackle two subtasks: (T1) recovering user routes (trajectory reconstruction); and (T2) predicting the trip that the user would travel (trajectory prediction). We show that the encoded contextual trajectory vectors efficiently characterize the hierarchical mobility semantics, from which one can decode the implicit meanings of trajectories. We evaluate our method on several public datasets and demonstrate that the proposed CATHI can efficiently improve the performance of both subtasks, compared to state-of-the-art approaches.

Point-of-interest (POI) recommendation is essential to a variety of services for both users and business. An extensive number of models have been developed to improve the recommendation performance by exploiting various characteristics and relations among POIs (e.g., spatio-temporal, social, etc.). However, very few studies closely look into the underlying mechanism accounting for why users prefer certain POIs to others. In this work, we initiate the first attempt to learn the distribution of user latent preference by proposing an Adversarial POI Recommendation (APOIR) model, consisting of two major components: (1) the recommender (R) which suggests POIs based on the learned distribution by maximizing the probabilities that these POIs are predicted as unvisited and potentially interested; and (2) the discriminator (D) which distinguishes the recommended POIs from the true check-ins and provides gradients as the guidance to improve R in a rewarding framework. Two components are co-trained by playing a minimax game towards improving itself while pushing the other to the boundary. By further integrating geographical and social relations among POIs into the reward function as well as optimizing R in a reinforcement learning manner, APOIR obtains significant performance improvement in four standard metrics compared to the state of the art methods.

An important task in Location based Social Network applications is to predict mobility - specifically, user's next point-of-interest (POI) - challenging due to the implicit feedback of footprints, sparsity of generated check-ins, and the joint impact of historical periodicity and recent check-ins. Motivated by recent success of deep variational inference, we propose VANext (Variational Attention based Next) POI prediction: a latent variable model for inferring user's next footprint, with historical mobility attention. The variational encoding captures latent features of recent mobility, followed by searching the similar historical trajectories for periodical patterns. A trajectory convolutional network is then used to learn historical mobility, significantly improving the efficiency over often used recurrent networks. A novel variational attention mechanism is proposed to exploit the periodicity of historical mobility patterns, combined with recent check-in preference to predict next POIs. We also implement a semi-supervised variant - VANext-S, which relies on variational encoding for pre-training all current trajectories in an unsupervised manner, and uses the latent variables to initialize the current trajectory learning. Experiments conducted on real-world datasets demonstrate that VANext and VANext-S outperform the state-of-the-art human mobility prediction models.

Recent advances in network representation learning have enabled significant improvements in the link prediction task, which is at the core of many downstream applications. As an increasing amount of mobility data becoming available due to the development of location technologies, we argue that this resourceful user mobility data can be used to improve link prediction performance. In this paper, we propose a novel link prediction framework that utilizes user offline check-in behavior combined with user online social relations. We model user offline location preference via probabilistic factor model and represent user social relations using neural network embedding. Furthermore, we employ locality-sensitive hashing to project the aggregated user representation into a binary matrix, which not only preserves the data structure but also speeds up the followed convolutional network learning. By comparing with several baseline methods that solely rely on social network or mobility data, we show that our unified approach significantly improves the performance.

Learning explicit and implicit patterns in human trajectories plays an important role in many Location-Based Social Networks (LBSNs) applications, such as trajectory classification (e.g., walking, driving, etc.), trajectory-user linking, friend recommendation, etc. A particular problem that has attracted much attention recently - and is the focus of our work - is the Trajectory-based Social Circle Inference (TSCI), aiming at inferring user social circles (mainly social friendship) based on motion trajectories and without any explicit social networked information. Existing approaches addressing TSCI lack satisfactory results due to the challenges related to data sparsity, accessibility and model efficiency. Motivated by the recent success of machine learning in trajectory mining, in this paper we formulate TSCI as a novel multi-label classification problem and develop a Recurrent Neural Network (RNN)-based framework called DeepTSCI to use human mobility patterns for inferring corresponding social circles. We propose three methods to learn the latent representations of trajectories, based on: (1) bidirectional Long Short-Term Memory (LSTM); (2) Autoencoder; and (3) Variational autoencoder. Experiments conducted on real-world datasets demonstrate that our proposed methods perform well and achieve significant improvement in terms of macro-R, macro-F1 and accuracy when compared to baselines.

Trajectory-User Linking (TUL) is an essential task in Geo-tagged social media (GTSM) applications, enabling personalized Point of Interest (POI) recommendation and activity identification. Existing works on mining mobility patterns often model trajectories using Markov Chains (MC) or recurrent neural networks (RNN) -- either assuming independence between non-adjacent locations or following a shallow generation process. However, most of them ignore the fact that human trajectories are often sparse, high-dimensional and may contain embedded hierarchical structures. We tackle the TUL problem with a semi-supervised learning framework, called TULVAE (TUL via Variational AutoEncoder), which learns the human mobility in a neural generative architecture with stochastic latent variables that span hidden states in RNN. TULVAE alleviates the data sparsity problem by leveraging large-scale unlabeled data and represents the hierarchical and structural semantics of trajectories with high-dimensional latent variables. Our experiments demonstrate that TULVAE improves efficiency and linking performance in real GTSM datasets, in comparison to existing methods.

The typical aim of User Identity Linkage (UIL) is to detect when users from across different social platforms are actually one and the same individual. Existing efforts to address this problem of practical relevance span from user-profile-based, through user-generated-content-based, user-behavior-based approaches to supervised or unsupervised learning frameworks, to subspace learning-based models. Most of them often require extraction of relevant features (e.g., profile, location, biography, networks, behavior, etc.) to model the user consistently across different social networks. However, these features are mainly derived based on prior knowledge and may vary for different platforms and applications. Inspired by the recent successes of deep learning in different tasks, especially in automatic feature extraction and representation, we propose a deep neural network based algorithm for UIL, called DeepLink. It is a novel end-to-end approach in a semi-supervised learning manner, without involving any hand-crafting features. Specifically, DeepLink samples the networks and learns to encode network nodes into vector representation to capture local and global network structures which, in turn, can be used to align anchor nodes through deep neural networks. A dual learning based paradigm is exploited to learn how to transfer knowledge and update the linkage using the policy gradient method. Experiments conducted on several public datasets show that DeepLink outperforms the state-of-the-art methods in terms of both linking precision and identity-match ranking.