Text-attributed graphs (TAGs) enhance graph learning by integrating rich textual semantics and topological context for each node. While boosting expressiveness, they also expose new vulnerabilities in graph learning through text-based adversarial surfaces. Recent advances leverage diverse backbones, such as graph neural networks (GNNs) and pre-trained language models (PLMs), to capture both structural and textual information in TAGs. This diversity raises a key question: How can we design universal adversarial attacks that generalize across architectures to assess the security of TAG models? The challenge arises from the stark contrast in how different backbones—GNNs and PLMs—perceive and encode graph patterns, coupled with the fact that many PLMs are only accessible via APIs, limiting attacks to black-box settings. To address this, we propose BadGraph, a novel attack framework that deeply elicits large language models' (LLMs) understanding of general graph knowledge to jointly perturb both node topology and textual semantics. Specifically, we design a target influencer retrieval module that leverages graph priors to construct cross-modally aligned attack shortcuts, thereby enabling efficient LLM-based perturbation reasoning. Experiments show that BadGraph achieves universal and effective attacks across GNN- and LLM-based reasoners, with up to a 76.3% performance drop, while theoretical and empirical analyses confirm its stealthy yet interpretable nature.

MOTIVATION: Accurate prediction of adverse drug reactions (ADRs) is essential for drug safety surveillance, and recent advances in machine learning with heterogeneous biomedical information have improved predictive performance. However, two challenges remain: current methods often learn inadequate ADR representations that fail to capture dependencies among ADRs, and generalize poorly to novel drugs. RESULTS: To obtain informative ADR embeddings, we construct a multi-source, multi-relational ADR graph that integrates hierarchical structure and empirical ADR co-occurrence, and apply a relational graph convolutional network (R-GCN) to learn relation-aware ADR representations. To enhance generalization to novel drugs, we exploit the hierarchical structure of the Anatomical Therapeutic Chemical (ATC) classification to link drugs via shared higher-level categories for effective knowledge transfer and model these relations with an R-GCN. We further introduce a Conditional Domain Adversarial Network (CDAN) to reduce distribution shifts between known and novel drugs by aligning features conditioned on predicted ADR labels, learning domain-invariant yet task-relevant representations. Additionally, to exploit similar ADR patterns among related drugs, we introduce a dual-branch mixture-of-experts (Dual-MoE) module where each expert captures ADR commonalities within a drug category in one branch, while a separate branch models global patterns. Extensive experiments show that our method consistently outperforms seven baselines, achieving F1 improvements of 4.3% and 4.7% over the best baseline on two datasets, respectively, with more balanced precision-recall trade-offs. It also improves AUC on uncommon ADRs by 7% more than on common ADRs, and remains more robust under data sparsity, with more gradual performance degradation as training data decreases. AVAILABILITY AND IMPLEMENTATION: The code of our model is available at https://github.com/fzsdb/Knowledge-guided-ADR-prediction.git.

Thermal power plants rely on extensive sensor networks to monitor key operational parameters, yet harsh industrial environments often lead to incomplete data characterized by significant noise, complex physical dependencies, and abrupt state transitions. This impedes accurate monitoring and predictive analyses. To address these domain-specific challenges, we propose a novel Hybrid Multi-Head Attention (HybridMHA) model for time series imputation. The core novelty of our approach lies in the synergistic combination of diagonally-masked self-attention and dynamic sparse attention. Specifically, the diagonally-masked component strictly preserves temporal causality to model the sequential evolution of plant states, while the dynamic sparse component selectively identifies critical cross-variable dependencies, effectively filtering out sensor noise. This tailored design enables the model to robustly capture sparse physical inter-dependencies even during abrupt operational shifts.Using a real-world dataset from a thermal power plant, our model demonstrates statistically significant improvements, outperforming existing methods by 10%–20% on key metrics. Further validation on a public benchmark dataset confirms its generalizability. These findings highlight the model's potential for robust real-time monitoring in complex industrial applications. The code is publicly available at:https://github.com/huangls2024/ImputationTSD.

Emotion recognition using electroencephalogram (EEG) signals has broad potential across various domains. EEG signals have ability to capture rich spatial information related to brain activity, yet effectively modeling and utilizing these spatial relationships remains a challenge. Existing methods struggle with simplistic spatial structure modeling, failing to capture complex node interactions, and lack generalizable spatial connection representations, failing to balance the dynamic nature of brain networks with the need for discriminative and generalizable features. To address these challenges, we propose the Multi-granularity Integration Network with Discrete Codebook for EEG-based Emotion Recognition (MIND-EEG). The framework employs a multi-granularity approach, integrating global and regional spatial information through a Global State Encoder, an Intra-Regional Functionality Encoder, and an Inter-Regional Interaction Encoder to comprehensively model brain activity. Additionally, we introduce a discrete codebook mechanism for constructing network structures via vector quantization, ensuring compact and meaningful brain network representations while mitigating over-smoothing and enhancing model generalization. The proposed framework effectively captures the dynamic and diverse nature of EEG signals, enabling robust emotion recognition. Extensive comparisons and analyses demonstrate the effectiveness of MIND-EEG, and the source code is publicly available athttps://github.com/XJTU-EEG/MIND_EEG.

Attention is a vital cognitive process in the learning and memory environment, particularly in the context of online learning. Traditional methods for classifying attention states of online learners based on behavioral signals are prone to distortion, leading to increased interest in using electroencephalography (EEG) signals for authentic and accurate assessment. However, the field of attention state classification based on EEG signals in online learning faces challenges, including the scarcity of publicly available datasets, the lack of standardized data collection paradigms, and the requirement to consider the interplay between attention and other psychological states. In light of this, we present the Multi-label EEG dataset for classifying Mental Attention states (MEMA) in online learning. We meticulously designed a reliable and standard experimental paradigm with three attention states: neutral, relaxing, and concentrating, considering human physiological and psychological characteristics. This paradigm collected EEG signals from 20 subjects, each participating in 12 trials, resulting in 1,060 minutes of data. Emotional state labels, basic personal information, and personality traits were also collected to investigate the relation-ship between attention and other psychological states. Extensive quantitative and qualitative analysis, including a multi-label correlation study, validated the quality of the EEG attention data. The MEMA dataset and analysis provide valuable insights for advancing research on attention in online learning. The dataset is publicly available at https://github.com/XJTU-EEG/MEMA.

Multivariate Time Series Forecasting (MTSF) has long been a key research focus. Traditionally, these studies assume a fixed number of variables, but in real-world applications, Cyber-Physical Systems often expand as new sensors are deployed, increasing variables in MTSF. In light of this, we introduce a novel task, Expanding-variate Time Series Forecasting (EVTSF). This task presents unique challenges, specifically (1) handling inconsistent data shapes caused by adding new variables, and (2) addressing imbalanced spatio-temporal learning, where expanding variables have limited observed data due to the necessity for timely operation. To address these challenges, we propose STEV, a flexible spatio-temporal forecasting framework. STEV includes a new Flat Scheme to tackle the inconsistent data shape issue, which extends the graph-based spatio-temporal modeling architecture into 1D space by flattening the 2D samples along the variable dimension, making the model variable-scale-agnostic while still preserving dynamic spatial correlations through a holistic graph. Additionally, we introduce a novel Spatio-temporal Focal Learning strategy that incorporates a negative filter to resolve potential conflicts between contrastive learning and graph representation, and a focal contrastive loss as its core to guide the framework to focus on optimizing the expanding variables. To evaluate the effectiveness of STEV, we benchmark EVTSF performance on three real-world datasets from various domains and compare it against three potential solutions employing state-of-the-art (SOTA) MTSF models tailored for EVSTF. Experimental results show that STEV significantly outperforms its competitors, especially in handling expanding variables. Notably, STEV, with only 5% of observations during the expanding period, is on par with SOTA MTSF models trained with complete data. Further exploration of various expanding scenarios underscores the generalizability of STEV in real-world applications.

Remote sensing change detection is essential for monitoring urban expansion, disaster assessment, and resource management, offering timely, accurate, and large-scale insights into dynamic landscape transformations. While deep learning has revolutionized change detection, the increasing complexity and computational demands of modern models have not necessarily translated into significant accuracy gains. Instead of following this trend, this study explores a more efficient approach, focusing on lightweight models that maintain high accuracy while minimizing resource consumption, which is an essential requirement for on-satellite processing. To this end, we propose FlickCD, which means quick flick then get great results, pushing the boundaries of the performance-resource trade-off. FlickCD introduces an Enhanced Difference Module (EDM) to amplify critical feature differences between temporal phases while suppressing irrelevant variations such as lighting and weather changes, thereby reducing computational costs in the subsequent change decoder. Additionally, the FlickCD decoder incorporates Local-Global Fusion Blocks, leveraging Shifted Window Self-Attention (SWSA) and Efficient Global Self-Attention (EGSA) to effectively capture semantic information at multiple scales, preserving both coarse- and fine-grained changes. Extensive experiments on four benchmark datasets demonstrate that FlickCD reduces computational and storage overheads by more than an order of magnitude while achieving state-of-the-art (SOTA) performance or incurring only a minor (<1% F1) accuracy trade-off. The implementation code is publicly available at https://github.com/xulsh8/FlickCD.

Time Series Forecasting (TSF) is a crucial task in various domains, yet existing TSF models rely heavily on high-quality data and insufficiently exploit all available data. This paper explores a novel self-supervised approach to re-label time series datasets by inherently constructing candidate datasets. During the optimization of a simple reconstruction network, intermediates are used as pseudo labels in a self-supervised paradigm, improving generalization for any predictor. We introduce the Self-Correction with Adaptive Mask (SCAM), which discards overfitted components and selectively replaces them with pseudo labels generated from reconstructions. Additionally, we incorporate Spectral Norm Regularization (SNR) to further suppress overfitting from a loss landscape perspective. Our experiments on eleven real-world datasets demonstrate that SCAM consistently improves the performance of various backbone models. This work offers a new perspective on constructing datasets and enhancing the generalization of TSF models through self-supervised learning. The code is available at https://github.com/SuDIS-ZJU/SCAM.

Decentralized Knowledge Graph (DKG) platforms allow the sharing of knowledge with multiple owners. While data owners can share their data with others by encrypting their data before sharing it, this naïve approach prevents data encrypted by different owners from being queried together, as it compromises query verifiability, an essential DKG platform feature. We propose Pistis, the first DKG platform capable of preserving ownership while also enabling verifiable SPARQL queries. Two novel techniques facilitate this: owner-managed end-to-end encryption and collaborative query verification. In Pistis, data owners thus encrypt their data individually and collaborate to construct an authenticated data structure (ADS) with a global key by means of secret sharing and secure multi-party computation. Then, by indexing KG data as ciphertext over the ADS, Pistis offers a cryptographic scheme called VO-SPARQL that facilitates verifiable queries on encrypted KG data with multiple owners. Pistis provides succinct proofs for two-stage SPARQL queries, including subgraph queries based on the ADS and aggregation on encrypted intermediate results based on a key-aggregate cryptographic primitive. A theoretical analysis and an empirical study provide detailed insight into the performance of Pistis while offering provable security.

EEG-based emotion recognition (EER) has gained significant attention due to its potential for understanding and analyzing human emotions. While recent advancements in deep learning techniques have substantially improved EER, the field lacks a convincing benchmark and comprehensive open-source libraries. This absence complicates fair comparisons between models and creates reproducibility challenges for practitioners, which collectively hinder progress. To address these issues, we introduce LibEER, a comprehensive benchmark and algorithm library designed to facilitate fair comparisons in EER. LibEER carefully selects popular and powerful baselines, harmonizes key implementation details across methods, and provides a standardized codebase in PyTorch. By offering a consistent evaluation framework with standardized experimental settings, LibEER enables unbiased assessments of seventeen representative deep learning models for EER across the six most widely used datasets. Additionally, we conduct a thorough, reproducible comparison of model performance and efficiency, providing valuable insights to guide researchers in the selection and design of EER models. Moreover, we make observations and in-depth analysis on the experiment results and identify current challenges in this community. We hope that our work will not only lower entry barriers for newcomers to EEG-based emotion recognition but also contribute to the standardization of research in this domain, fostering steady development. The library and source code are publicly available athttps://github.com/XJTU-EEG/LibEER.

Multi-unmanned aerial vehicle (UAV) cooperative maritime object recognition aims to maintain high accuracy under dynamic aerial perspectives for maritime search and rescue missions. Existing continual learning methods retain knowledge by approximating the global distribution of prior data but fail to address cross-perspective knowledge conflicts caused by distribution shifts across aerial perspectives, leading to gradient perturbations that harm consistency and accuracy in dynamic maritime environments. To address the issues, we propose a conflict-modulated generative continual learning (ConMod) framework, comprising generative perspective-robust conflict estimation and conflict-modulated continual learning modules. The generative perspective-robust conflict estimation employs a perspective-aware scene generator that embeds maritime knowledge priors as perspective constraints to augment the data distribution, thereby facilitating explainable cross-perspective conflict association and promoting robust conflict index estimation. It also incorporates a dual-modal conflict index estimator that integrates geometric distortion and environmental variation branches to estimate conflict indices by associating simulated scenes with perspective-robust data distributions. Furthermore, conflict-modulated continual learning introduces a perspective-specific triplet loss to regularize consistency by aligning geometric and environmental features within perspective-specific representation space. Additionally, a conflict-modulated loss treats conflict indices as modulation weights to identify and reinforce representative conflict experiences from a cross-perspective memory buffer, guiding gradient updates toward global optimization. Results on the SeaDronesSee-CL and SeaDronesSee-CL-v2 datasets show that ConMod effectively improves multi-UAV cooperative maritime object recognition by identifying and mitigating cross-perspective knowledge conflicts.

The critical importance of monitoring and recognizing human emotional states in healthcare has led to a surge in proposals for EEG-based multimodal emotion recognition in recent years. However, practical challenges arise in acquiring EEG signals in daily healthcare settings due to stringent data acquisition conditions, resulting in the issue of incomplete modalities. Existing studies have turned to knowledge distillation as a means to mitigate this problem by transferring knowledge from multimodal networks to unimodal ones. However, these methods are constrained by the use of a single teacher model to transfer integrated feature extraction knowledge, particularly concerning spatial and temporal features in EEG data. To address this limitation, we propose a multi-teacher knowledge distillation framework enhanced with a Large Language Model (LLM), aimed at facilitating effective feature learning in the student network by transferring knowledge of extracting integrated features. Specifically, we employ an LLM as the teacher for extracting temporal features and a graph convolutional neural network for extracting spatial features. To further enhance knowledge distillation, we introduce causal masking and a confidence indicator into the LLM to facilitate the transfer of the most discriminative features. Extensive testing on the DEAP and MAHNOB-HCI datasets demonstrates that our model outperforms existing methods in the modality-incomplete scenario. This study underscores the potential application of large models in this field.

Trajectory similarity computation serves as a fundamental functionality of various spatial information applications. Although existing deep learning similarity computation methods offer better efficiency and accuracy than non-learning solutions, they are still immature in trajectory embedding and suffer from poor generality and heavy preprocessing for training. Targeting these limitations, we propose a novel framework named KGTS based on knowledge graph grid embedding, prompt trajectory embedding, and unsupervised contrastive learning for improved trajectory similarity computation. Specifically, we first embed map grids with a GRot embedding method to vigorously grasp the neighbouring relations of grids. Then, a prompt trajectory embedding network incorporates the resulting grid embedding and extracts trajectory structure and point order information. It is trained by unsupervised contrastive learning, which not only alleviates the heavy preprocessing burden but also provides exceptional generality with creatively designed strategies for positive sample generation. The prompt trajectory embedding adopts a customized prompt paradigm to mitigate the gap between the grid embedding and the trajectory embedding. Extensive experiments on two real-world trajectory datasets demonstrate the superior performance of KGTS over state-of-the-art methods.

Imputation of Correlated Time Series (CTS) is essential in data preprocessing for many tasks, particularly when sensor data is often incomplete. Deep learning has enabled sophisticated models that improve CTS imputation by capturing temporal and spatial patterns. However, deep models often incur considerable consumption of computational resources and thus cannot be deployed in resource-limited settings. This paper presents ReCTSi (Resource-efficient CTS imputation), a method that adopts a new architecture for decoupled pattern learning in two phases: (1) the Persistent Pattern Extraction phase utilizes a multi-view learnable codebook mechanism to identify and archive persistent patterns common across different time series, enabling rapid pattern retrieval during inference. (2) the Transient Pattern Adaptation phase introduces completeness-aware attention modules that allocate attention to the complete and hence more reliable data segments. Extensive experimental results show that ReCTSi achieves state-of-the-art imputation accuracy while consuming much fewer computational resources than the leading existing model, consuming only 0.004% of the FLOPs for inference compared to its closest competitor. The blend of high accuracy and very low resource consumption makes ReCTSi the currently best method for resource-limited scenarios. The related code is available at https://github.com/ryanlaics/RECTSI.

Cyber-physical system sensors emit multivariate time series (MTS) that monitor physical system processes. Such time series generally capture unknown numbers of states, each with a different duration, that correspond to specific conditions, e.g., "walking" or "running" in human-activity monitoring. Unsupervised identification of such states facilitates storage and processing in subsequent data analyses, as well as enhances result interpretability. Existing state-detection proposals face three challenges. First, they introduce substantial computational overhead, rendering them impractical in resourceconstrained or streaming settings. Second, although state-of-the-art (SOTA) proposals employ contrastive learning for representation, insufficient attention to false negatives hampers model convergence and accuracy. Third, SOTA proposals predominantly only emphasize offline non-streaming deployment, we highlight an urgent need to optimize online streaming scenarios. We propose E2Usd that enables efficient-yet-accurate unsupervised MTS state detection. E2Usd exploits a Fast Fourier Transform-based Time Series Compressor (fftCompress) and a Decomposed Dual-view Embedding Module (ddEM) that together encode input MTSs at low computational overhead. Additionally, we propose a False Negative Cancellation Contrastive Learning method (fnccLearning) to counteract the effects of false negatives and to achieve more cluster-friendly embedding spaces. To reduce computational overhead further in streaming settings, we introduce Adaptive Threshold Detection (adaTD). Comprehensive experiments with six baselines and six datasets offer evidence that E2Usd is capable of SOTA accuracy at significantly reduced computational overhead. Our code is available at https://github.com/AI4CTS/E2Usd.

Societal and industrial infrastructures and systems increasingly leverage sensors that emit correlated time series. Forecasting of future values of such time series based on recorded historical values has important benefits. Automatically designed models achieve higher accuracy than manually designed models. Given a forecasting task, which includes a dataset and a forecasting horizon, automated design methods automatically search for an optimal forecasting model for the task in a manually designed search space, and then train the identified model using the dataset to enable the forecasting. Existing automated methods face three challenges. First, the search space is constructed by human experts, rending the methods only semi-automated and yielding search spaces prone to subjective biases. Second, it is time consuming to search for an optimal model. Third, training the identified model for a new task is also costly. These challenges limit the practicability of automated methods in real-world settings. To contend with the challenges, we propose a fully automated and highly efficient correlated time series forecasting framework where the search and training can be done in minutes. The framework includes a data-driven, iterative strategy to automatically prune a large search space to obtain a high-quality search space for a new forecasting task. It includes a zero-shot search strategy to efficiently identify the optimal model in the customized search space. And it includes a fast parameter adaptation strategy to accelerate the training of the identified model. Experiments on seven benchmark datasets offer evidence that the framework is capable of state-of-the-art accuracy and is much more efficient than existing methods.

Correlated time series (CTS) forecasting is essential in many practical applications, such as traffic management and server load control. Various deep learning based solutions have been proposed to improve forecasting accuracy. However, while models have become increasingly computationally intensive, they struggle to improve accuracy. This study aims instead to enable more lightweight, accurate models suitable for resource-constrained devices. To achieve this goal, we characterize popular CTS forecasting models, yielding two observations for developing lightweight CTS forecasting. On this basis, we propose theLightCTSframework that adopts plain stacking of temporal and spatial operators instead of alternate stacking which is much more computationally expensive. Moreover,LightCTSfeatures light temporal and spatial operators, L-TCN and GL-Former, offering improved computational efficiency without compromising their feature extraction capabilities.LightCTSalso encompasses a last-shot compression scheme to reduce redundant temporal features and speed up subsequent computations. Next, we equipLightCTSwith two knowledge distillation modules,TafdandCaad, that result inLightCTS$^\star$retaining the original benefits ofLightCTS, while also being able to adapt to varying levels of ultra-constrained resources. Experimental studies offer detailed insight into these proposals and provide evidence that bothLightCTSandLightCTS$^\star$are capable of nearly state-of-the-art accuracy at substantially reduced computational costs.

Uncertainty is an essential consideration for time series forecasting tasks. In this work, we focus on quantifying the uncertainty of traffic forecasting from a unified perspective. We develop a novel traffic forecasting framework, namely Deep Spatio-Temporal Uncertainty Quantification (DeepSTUQ), which can estimate both aleatoric and epistemic uncertainty. Specifically, we first leverage a spatio-temporal model to model the complex spatio-temporal correlations of traffic data. Subsequently, two independent sub-neural networks maximizing the heterogeneous log-likelihood are developed to estimate aleatoric uncertainty. To estimate epistemic uncertainty, we combine the merits of variational inference and deep ensembling by integrating the Monte Carlo dropout and the Adaptive Weight Averaging re-training methods, respectively. Furthermore, to relax the Gaussianity assumption, mitigate overfitting, and improve horizon-wise uncertainty quantification performance, we define a new calibration method called Multi-horizon Conformal Calibration (MHCC). Finally, we provide a theoretical analysis of the proposed unified approach based on the PAC-Bayes theory. Extensive experiments are conducted on four public datasets, and the empirical results suggest that the proposed method outperforms state-of-the-art methods in terms of both point prediction and uncertainty quantification.

Emotion recognition based on electroencephalography (EEG) is becoming a growing concern of researchers due to its various applications and portable devices. Existing methods are mainly dedicated to EEG feature representation and have made impressive progress. However, the problem of scarce labels restricts their further promotion. In light of this, we propose a self-supervised framework with contrastive learning for robust EEG-based emotion recognition, which can effectively leverage both readily available unlabeled EEG signals and labeled ones to learn highly discriminative EEG features. Firstly, we construct a specific pretext task according to the sequential non-stationarity of emotional EEG signals for contrastive learning, which aims at extracting pseudo-label information from all EEG data. Meanwhile, we propose a novel negative segment selection algorithm to reduce the noise of unlabeled data during the contrastive learning process. Secondly, to mitigate the overfitting issue induced by a small number of labeled samples during learning, we originate a loss function with label smoothing regularization that can guide the model to learn generalizable features. Extensive experiments over three benchmark datasets demonstrate the effectiveness and superiority of our model on EEG-based emotion recognition task. Besides, the generalization and robustness of the model have also been proved through sufficient experiments.

Uncertainty is an essential consideration for time series forecasting tasks. In this work, we specifically focus on quantifying the uncertainty of traffic forecasting. To achieve this, we develop Deep Spatio-Temporal Uncertainty Quantification (DeepSTUQ), which can estimate both aleatoric and epistemic uncertainty. We first leverage a spatio-temporal model to model the complex spatio-temporal correlations of traffic data. Subsequently, two independent sub-neural networks maximizing the heterogeneous log-likelihood are developed to estimate aleatoric uncertainty. For estimating epistemic uncertainty, we combine the merits of variational inference and deep ensembling by integrating the Monte Carlo dropout and the Adaptive Weight Averaging re-training methods, respectively. Finally, we propose a post-processing calibration approach based on Temperature Scaling, which improves the model’s generalization ability to estimate uncertainty. Extensive experiments are conducted on four public datasets, and the empirical results suggest that the proposed method outperforms state-of-the-art methods in terms of both point prediction and uncertainty quantification.

Electroencephalogram (EEG)-based brain–computer interfaces (BCIs) allow users to use brain signals to control external instruments, and movement intention detecting BCIs can aid in the rehabilitation of patients who have lost motor function. Existing studies in this area mostly rely on cue-based data collection that facilitates sample labeling but introduces noise from cue stimuli; moreover, it requires extensive user training, and cannot reflect real usage scenarios. In contrast, self-paced BCIs can overcome the limitations of the cue-based approach by supporting users to perform movements at their own initiative and pace, but they fall short in labeling. Therefore, in this study, we proposed an automated labeling approach that can cross-reference electromyography (EMG) signals for EEG labeling with zero human effort. Furthermore, considering that only a few studies have focused on evaluating BCI systems for online use and most of them do not report details of the online systems, we developed and present in detail a pseudo-online evaluation suite to facilitate online BCI research. We collected self-paced movement EEG data from 10 participants performing opening and closing hand movements for training and evaluation. The results show that the automated labeling method can contend well with noisy data compared with the baseline labeling method. We also explored popular machine learning models for online self-paced movement detection. The results demonstrate the capability of our online pipeline, and that a well-performing offline model does not necessarily translate to a well-performing online model owing to the specific settings of an online BCI system. Our proposed automated labeling method, online evaluation suite, and dataset take a concrete step towards real-world self-paced BCI systems.

Correlated time series (CTS) forecasting plays an essential role in many practical applications, such as traffic management and server load control. Many deep learning models have been proposed to improve the accuracy of CTS forecasting. However, while models have become increasingly complex and computationally intensive, they struggle to improve accuracy. Pursuing a different direction, this study aims instead to enable much more efficient, lightweight models that preserve accuracy while being able to be deployed on resource-constrained devices. To achieve this goal, we characterize popular CTS forecasting models and yield two observations that indicate directions for lightweight CTS forecasting. On this basis, we propose the LightCTS framework that adopts plain stacking of temporal and spatial operators instead of alternate stacking that is much more computationally expensive. Moreover, LightCTS features light temporal and spatial operator modules, called L-TCN and GL-Former, that offer improved computational efficiency without compromising their feature extraction capabilities. LightCTS also encompasses a last-shot compression scheme to reduce redundant temporal features and speed up subsequent computations. Experiments with single-step and multi-step forecasting benchmark datasets show that LightCTS is capable of nearly state-of-the-art accuracy at much reduced computational and storage overheads.

Sensors in cyber-physical systems often capture interconnected processes and thus emit correlated time series (CTS), the forecasting of which enables important applications. The key to successful CTS forecasting is to uncover the temporal dynamics of time series and the spatial correlations among time series. Deep learning-based solutions exhibit impressive performance at discerning these aspects. In particular, automated CTS forecasting, where the design of an optimal deep learning architecture is automated, enables forecasting accuracy that surpasses what has been achieved by manual approaches. However, automated CTS solutions remain in their infancy and are only able to find optimal architectures for predefined hyperparameters and scale poorly to large-scale CTS. To overcome these limitations, we propose AutoCTS+, a joint, scalable framework, to automatically devise effective CTS forecasting models. Specifically, we encode each candidate architecture and accompanying hyperparameters into a joint graph representation. We introduce an efficient Architecture-Hyperparameter Comparator (AHC) to rank all architecture-hyperparameter pairs, and we then further evaluate the top-ranked pairs to select an architecture-hyperparameter pair as the final model. Extensive experiments on six benchmark datasets demonstrate that AutoCTS+ not only eliminates manual efforts but also is capable of better performance than manually designed and existing automatically designed CTS models. In addition, it shows excellent scalability to large CTS.

The ability to decentralize knowledge graphs (KG) is important to exploit the full potential of the Semantic Web and realize the Web 3.0 vision. However, decentralization also renders KGs more prone to attacks with adverse effects on data integrity and query verifiability. While existing studies focus on ensuring data integrity, how to ensure query verifiability - thus guarding against incorrect, incomplete, or outdated query results - remains unsolved. We propose VeriDKG, the first SPARQL query engine for decentralized knowledge graphs (DKG) that offers both data integrity and query verifiability guarantees. The core of VeriDKG is the RGB-Trie, a new blockchain-maintained authenticated data structure (ADS) facilitating correctness proofs for SPARQL query results. VeriDKG enables verifiability of subqueries by gathering global index information on subgraphs using the RGB-Trie, which is implemented as a new variant of the Merkle prefix tree with an RGB color model. To enable verifiability of the final query result, the RGB-Trie is integrated with a cryptographic accumulator to support verifiable aggregation operations. A rigorous analysis of query verifiability in VeriDKG is presented, along with evidence from an extensive experimental study demonstrating its state-of-the-art query performance on the largeRDFbench benchmark.

Emotion recognition based on electroencephalography (EEG) has attracted significant attention due to its wide range of applications, especially in Human-Computer Interaction(HCI). Previous research treats different segments of EEG signals uniformly, ignoring the fact that emotions are unstable and discrete during an extended period. In this paper, we propose a novel two-step spatial-temporal emotion recognition framework. First, considering that the human emotion has not only ”short-term continuity” but also ”long-term similarity”, we propose a hierarchical self-attention network to jointly model local and global temporal information, so as to localize most related segments and reduce the influence of noise at the temporal level. Second, in order to extract discriminative features at the spatial level to enhance the emotion recognition performance, we further employ the squeeze-and-excitation module (SE module) along with the channel correlation loss (CC-Loss) to select the most task-related channels. We also define a new task calledemotion localization, which aims to localize fragments with stronger emotions. We evaluate the proposed method on the proposed emotion localization task and typical emotion recognition task with three publicly available datasets, i.e., SEED, DEAP, and MAHNOB-HCI. The experimental results demonstrate that the proposed approach outperforms state-of-the-art methods.

With the diffusion of online mobile devices with geo-location capabilities, the infrastructure necessary for real-world deployment of Spatial Crowdsourcing (SC), where so-called mobile workers are assigned location-sensitive tasks, is in place. Some SC tasks cannot be completed by a single worker due to their complexity, but rather must be assigned to and completed by a group of users. Achieving such group assignments that satisfy all group members evenly is an open challenge. To address this challenge, we propose a novel preference-aware group task assignment framework encompassing two components: Mutual Information-based Preference Modeling (MIPM) and Preference-aware Group Task Assignment (PGTA). The MIPM component learns the preferences of groups contrastively by maximizing the mutual information between workers and worker groups based on worker-task and group-task interaction data and by using an attention mechanism to weight group members adaptively. In addition, curriculum negative sampling is adopted to generate a small number of negative workers for each worker group, following the principles of curriculum learning. Next, the PGTA component offers an optimal task assignment algorithm that employs tree decomposition to assign tasks to appropriate worker groups, with the aim of maximizing the number of task assignments while prioritizing more interested groups when assigning tasks. The task assignment framework also features preference-constrained pruning of unpromising worker groups to speed up the assignment process. Finally, we report extensive experiments that offer evidence of the effectiveness and practicality of the paper's proposal.

Ubiquitous mobile sensors on human activity recognition pose the threat of leaking personal information that is implicitly contained within the time-series sensor signals and can be extracted by attackers. Existing protective methods only support specific sensitive attributes and require massive relevant sensitive ground truth for training, which is unfavourable to users. To fill this gap, we propose a novel data transformation framework for prohibiting the leakage of sensitive information from sensor data. The proposed framework transforms raw sensor data into a new format, where the sensitive information is hidden and the desired information (e.g., human activities) is retained. Training can be conducted without using any personal information as ground truth. Meanwhile, multiple attributes of sensitive information (e.g., age, gender) can be collectively hidden through a one-time transformation. The experimental results on two multimodal sensor-based human activity datasets manifest the feasibility of the presented framework in hiding users’ sensitive information (inference MAE increases$\sim$2 times and inference accuracy degrades$\sim$50%) without degrading the usability of the data for activity recognition (only$\sim$2% accuracy degradation).

Correlated time series (CTS) forecasting plays an essential role in many cyber-physical systems, where multiple sensors emit time series that capture interconnected processes. Solutions based on deep learning that deliver state-of-the-art CTS forecasting performance employ a variety of spatio-temporal (ST) blocks that are able to model temporal dependencies and spatial correlations among time series. However, two challenges remain. First, ST-blocks are designed manually, which is time consuming and costly. Second, existing forecasting models simply stack the same ST-blocks multiple times, which limits the model potential. To address these challenges, we propose AutoCTS that is able to automatically identify highly competitive ST-blocks as well as forecasting models with heterogeneous ST-blocks connected using diverse topologies, as opposed to the same ST-blocks connected using simple stacking. Specifically, we design both a micro and a macro search space to model possible architectures of ST-blocks and the connections among heterogeneous ST-blocks, and we provide a search strategy that is able to jointly explore the search spaces to identify optimal forecasting models. Extensive experiments on eight commonly used CTS forecasting benchmark datasets justify our design choices and demonstrate that AutoCTS is capable of automatically discovering forecasting models that outperform state-of-the-art human-designed models.

Brain-computer interface (BCI) is a system empowering humans to communicate with or control the outside world with exclusively brain intentions. Electroencephalography (EEG)-based BCI is one of the promising solutions due to its convenient and portable instruments. Despite the extensive research of EEG in recent years, it is still challenging to interpret EEG signals effectively due to its nature of noise and difficulties in capturing the inconspicuous relations between EEG signals and specific brain activities. Most existing works either only consider EEG as chain-like sequences while neglecting complex dependencies between adjacent signals or requiring complex preprocessing. In this paper, we introduce two deep learning-based frameworks with novel spatio-temporal preserving representations of raw EEG streams to precisely identify human intentions. The two frameworks consist of both convolutional and recurrent neural networks effectively exploring the preserved spatial and temporal information in either a cascade or a parallel manner. Extensive experiments on a large scale movement intention EEG dataset (108 subjects, 3 145 160 EEG records) have demonstrated that the proposed frameworks achieve high accuracy of 98.3% and outperform a set of state-of-the-art and baseline models. The developed models are further evaluated with a real-world brain typing BCI and achieve a recognition accuracy of 93% over five instruction intentions suggesting good generalization over different kinds of intentions and BCI systems.

Motor imagery classification from EEG signals is essential for motor rehabilitation with a Brain-Computer Interface (BCI). Most current works on this issue require a subject-specific adaptation step before applied to a new user. Thus the research of directly extending a pre-trained model to new users is particularly desired and indispensable. As brain dynamics fluctuate considerably across different subjects, it is challenging to design practical hand-crafted features based on prior knowledge. Regarding this gap, this paper proposes a Graph-based Convolutional Recurrent Attention Model (G-CRAM) to explore EEG features across different subjects for motor imagery classification. A graph structure is first developed to represent the positioning information of EEG nodes. Then a convolutional recurrent attention model learns EEG features from both spatial and temporal dimensions and emphasizes on the most distinguishable temporal periods. We evaluate the proposed approach on two benchmark EEG datasets of motor imagery classification on the subject-independent testing. The results show that the G-CRAM achieves superior performance to state-of-the-art methods regarding recognition accuracy and ROC-AUC. Furthermore, model interpretation studies reveal the learning process of different neural network components and demonstrate that the proposed model can extract detailed features efficiently.

Recent years have witnessed the success of deep learning methods in human activity recognition (HAR). The longstanding shortage of labeled activity data inherently calls for a plethora of semisupervised learning methods, and one of the most challenging and common issues with semisupervised learning is the imbalanced distribution of labeled data over classes. Although the problem has long existed in broad real-world HAR applications, it is rarely explored in the literature. In this paper, we propose a semisupervised deep model for imbalanced activity recognition from multimodal wearable sensory data. We aim to address not only the challenges of multimodal sensor data (e.g., interperson variability and interclass similarity) but also the limited labeled data and class-imbalance issues simultaneously. In particular, we propose a pattern-balanced semisupervised framework to extract and preserve diverse latent patterns of activities. Furthermore, we exploit the independence of multi-modalities of sensory data and attentively identify salient regions that are indicative of human activities from inputs by our recurrent convolutional attention networks. Our experimental results demonstrate that the proposed model achieves a competitive performance compared to a multitude of state-of-the-art methods, both semisupervised and supervised ones, with 10% labeled training data. The results also show the robustness of our method over imbalanced, small training data sets.

Semi-supervised learning is crucial for alleviating labelling burdens in people-centric sensing. However, humangenerated data inherently suffer from distribution shift in semi-supervised learning due to the diverse biological conditions and behavior patterns of humans. To address this problem, we propose a generic distributionally robust model for semi-supervised learning on distributionally shifted data. Considering both the discrepancy and the consistency between the labeled data and the unlabeled data, we learn the latent features that reduce person-specific discrepancy and preserve task-specific consistency. We evaluate our model in a variety of people-centric recognition tasks on real-world datasets, including intention recognition, activity recognition, muscular movement recognition and gesture recognition. The experiment results demonstrate that the proposed model outperforms the state-of-the-art methods.

As brain dynamics fluctuate considerably across different subjects, it is challenging to design effective handcrafted features based on prior knowledge. Regarding this gap, this paper proposes a Graph-based Convolutional Recurrent Attention Model (G-CRAM) to explore EEG features across different subjects for movement intention recognition. A graph structure is first developed to embed the positioning information of EEG nodes, and then a convolutional recurrent attention model learns EEG features from both spatial and temporal dimensions and adaptively emphasizes on the most distinguishable temporal periods. The proposed approach is validated on two public movement intention EEG datasets. The results show that the GCRAM achieves superior performance to state-of-the-art methods regarding recognition accuracy and ROC-AUC. Furthermore, model interpreting studies reveal the learning process of different neural network components and demonstrate that the proposed model can extract detailed features efficiently.

Sharing ubiquitous mobile sensor data, especially physiological data, raises potential risks of leaking physical and demographic information that can be inferred from the time series sensor data. Existing sensitive information protection mechanisms that depend on data transformation are effective only on a particular sensitive attribute, together with usually requiring the labels of sensitive information for training. Considering this gap, we propose a novel user sensitive information protection framework without using a sensitive training dataset or being validated on protecting only one specific sensitive information. The presented approach transforms raw sensor data into a new format that has a "style" (sensitive information) of random noise and a "content" (desired information) of the raw sensor data, thus is free of user sensitive information for training and able to collectively protect all sensitive information at once. Our implementation and experiments on two real-world multisensor human activity datasets demonstrate that the proposed data transformation technique can achieve the protection for all sensitive information at once without requiring the knowledge of users' personal attributes for training, and simultaneously preserve the usability of the new transformed data with regard to inferring human activities with insignificant performance loss.

Multi-modality is an important feature of sensor based activity recognition. In this work, we consider two inherent characteristics of human activities, the spatially-temporally varying salience of features and the relations between activities and corresponding body part motions. Based on these, we propose a multi-agent spatial-temporal attention model. The spatial-temporal attention mechanism helps intelligently select informative modalities and their active periods. And the multiple agents in the proposed model represent activities with collective motions across body parts by independently selecting modalities associated with single motions. With a joint recognition goal, the agents share gained information and coordinate their selection policies to learn the optimal recognition model. The experimental results on four real-world datasets demonstrate that the proposed model outperforms the state-of-the-art methods.

The electroencephalogram (EEG) signal is a medium to realize a brain-computer interface (BCI) system due to its zero clinical risk and portable acquisition devices. Current EEG-based BCI research usually requires a subject-specific adaptation step before a BCI can be employed by a new user. In contrast, the subject-independent scenario, where a well trained model can be directly applied to new users without precalibration, is particularly desired. Considering this critical gap, the focus in this letter is developing an effective EEG signal analysis adaptively applied to subject-independent settings. We present a convolutional recurrent attention model (CRAM) that utilizes a convolutional neural network to encode the high-level representation of EEG signals and a recurrent attention mechanism to explore the temporal dynamics of the EEG signals as well as to focus on the most discriminative temporal periods. Extensive experiments on a benchmark multiclass EEG dataset containing four movement intentions indicate that the proposed model is capable of exploiting the underlying invariant EEG patterns across different subjects and generalizing the patterns to new subjects with better performance than a series of state-of-the-art and baseline approaches by at least eight percentage points. The implementation code is made publicly available.11https://github.com/dalinzhang/CRAM.

Brain-Computer Interface (BCI) is a system empowering humans to communicate with or control the outside world with exclusively brain intentions. Electroencephalography (EEG) based BCIs are promising solutions due to their convenient and portable instruments. Despite the extensive research of EEG in recent years, it is still challenging to interpret EEG signals effectively due to the massive noises in EEG signals (e.g., low signal-noise ratio and incomplete EEG signals), and difficulties in capturing the inconspicuous relationships between EEG signals and certain brain activities. Most existing works either only consider EEG as chain-like sequences neglecting complex dependencies between adjacent signals or requiring pre-processing such as transforming EEG waves into images. In this paper, we introduce both cascade and parallel convolutional recurrent neural network models for precisely identifying human intended movements and instructions effectively learning the compositional spatio-temporal representations of raw EEG streams. Extensive experiments on a large scale movement intention EEG dataset (108 subjects,3,145,160 EEG records) have demonstrated that both models achieve high accuracy near 98.3% and outperform a set of baseline methods and most recent deep learning based EEG recognition models, yielding a significant accuracy increase of 18% in the cross-subject validation scenario. The developed models are further evaluated with a real-world BCI and achieve a recognition accuracy of 93% over five instruction intentions. This suggests the proposed models are able to generalize over different kinds of intentions and BCI systems.

Brain-Computer Interface (BCI) enables human to communicate with and intuitively control an external device through brain signals. Movement intention recognition paves the path for developing BCI applications. The current state-of-the-art in EEG based BCI usually involves subject-specific adaptation before ready to use. However, the subject-independent scenario, in which a well-trained model is directly applied to new subjects without any pre-calibration, is particularly desired yet rarely explored. In order to fill the gap, we present a Convolutional Attention Model (CAM) for EEG-based human movement intention recognition in the subject-independent scenario. The convolutional network is designed to capture the spatio-temporal features of EEG signals, while the integrated attention mechanism is utilized to focus on the most discriminative information of EEG signals during the period of movement imagination while omitting other less relative parts. Experiments conducted on a real-world EEG dataset containing 55 subjects show that our model is capable of mining the underlying invariant EEG patterns across different subjects and generalizing to unseen subjects. Our model achieves better performance than a series of state-of-the-art and baseline approaches.

Multimodal features play a key role in wearable sensor based human activity recognition (HAR). Selecting the most salient features adaptively is a promising way to maximize the effectiveness of multimodal sensor data. In this regard, we propose a “collect fully and select wisely” principle as well as an interpretable parallel recurrent model with convolutional attentions to improve the recognition performance. We first collect modality features and the relations between each pair of features to generate activity frames, and then introduce an attention mechanism to select the most prominent regions from activity frames precisely. The selected frames not only maximize the utilization of valid features but also reduce the number of features to be computed effectively. We further analyze the accuracy and interpretability of the proposed model based on extensive experiments. The results show that our model achieves competitive performance on two benchmarked datasets and works well in real life scenarios.

An electroencephalography (EEG) based Brain Computer Interface (BCI) enables people to communicate with the outside world by interpreting the EEG signals of their brains to interact with devices such as wheelchairs and intelligent robots. More specifically, motor imagery EEG (MI-EEG), which reflects a subject's active intent, is attracting increasing attention for a variety of BCI applications. Accurate classification of MI-EEG signals while essential for effective operation of BCI systems is challenging due to the significant noise inherent in the signals and the lack of informative correlation between the signals and brain activities. In this paper, we propose a novel deep neural network based learning framework that affords perceptive insights into the relationship between the MI-EEG data and brain activities. We design a joint convolutional recurrent neural network that simultaneously learns robust high-level feature presentations through low-dimensional dense embeddings from raw MI-EEG signals. We also employ an Autoencoder layer to eliminate various artifacts such as background activities. The proposed approach has been evaluated extensively on a large-scale public MI-EEG dataset and a limited but easy-to-deploy dataset collected in our lab. The results show that our approach outperforms a series of baselines and the competitive state-of-the-art methods, yielding a classification accuracy of 95.53%. The applicability of our proposed approach is further demonstrated with a practical BCI system for typing.

An electroencephalography (EEG) based brain activity recognition is a fundamental field of study for a number of significant applications such as intention prediction, appliance control, and neurological disease diagnosis in smart home and smart healthcare domains. Existing techniques mostly focus on binary brain activity recognition for a single person, which limits their deployment in wider and complex practical scenarios. Therefore, multi-person and multi-class brain activity recognition has obtained popularity recently. Another challenge faced by brain activity recognition is the low recognition accuracy due to the massive noises and the low signal-to-noise ratio in EEG signals. Moreover, the feature engineering in EEG processing is time-consuming and highly relies on the expert experience. In this paper, we attempt to solve the above challenges by proposing an approach which has better EEG interpretation ability via raw Electroencephalography (EEG) signal analysis for multi-person and multi-class brain activity recognition. Specifically, we analyze inter-class and inter-person EEG signal characteristics, based on which to capture the discrepancy of inter-class EEG data. Then, we adopt an Autoencoder layer to automatically refine the raw EEG signals by eliminating various artifacts. We evaluate our approach on both a public and a local EEG datasets and conduct extensive experiments to explore the effect of several factors (such as normalization methods, training data size, and Autoencoder hidden neuron size) on the recognition results. The experimental results show that our approach achieves a high accuracy comparing to competitive state-of-the-art methods, indicating its potential in promoting future research on multi-person EEG recognition.