Qi Zhu

BACKGROUND: Intermittent Valsalva maneuver (VM) is commonly used to facilitate high-intensity focused ultrasound (HIFU) ablation. However, the optimal duration of a VM in terms of adverse impacts on brain function is unknown. This prospective study explored the impact of different VM durations on brain function in patients undergoing HIFU ablation surgery for hepatic cancer. METHODS: Adult patients scheduled for HIFU ablation for liver cancer under general anesthesia were randomized into a control group, ≤5.2 min per VM episode (short VM group), or >5.2 min per VM episode (long VM group). A four-channel electroencephalogram was conducted using SedLine to record brain electrical parameters (e.g., power spectral density, entropy values, and cross-frequency domain coupling). The primary endpoint was burst suppression in the intent-to-treat population that included all enrolled subjects. Key secondary endpoints included postoperative delirium, emergence agitation, and postoperative recovery quality. RESULTS: A total of 156 subjects were screened, and 153 were randomized: 50, 52, and 51 subjects in the control, short VM, and long VM groups, respectively. The rate of burst suppression was 2.0% (1/50), 30.8% (16/52), and 66.7% (34/51) in the control, short VM, and long VM groups, respectively ( P < 0.001). The long VM group also had lower alpha and beta band power spectral density, and higher permutation entropy compared to the control group. The rate of postoperative delirium was 0.0% (0/48), 0.0% (0/49), and 6.1% (3/49) in the control, short VM, and long VM groups, respectively ( P = 0.107). The long VM group had the highest rate of emergence agitation (64.7% vs 2.0% in the control group and 28.8% in the short VM group, P < 0.001), and the lowest Quality of Recovery-15 scores [106 (100-109) vs 110 (107-118) in the control group and 110 (102-114) in the short VM group, P < 0.001]. CONCLUSION: A VM lasting for >5.2 min per episode resulted in substantial increases in the rates of burst suppression in adult patients undergoing HIFU ablation for liver cancer. Based on these findings, it is recommended that the VM should be limited to ≤5.2 min per episode if possible.

Imaging mass cytometry (IMC) enables the high-resolution spatial profiling of tumor microenvironments, but its clinical utility for prospective prediction remains underdeveloped. In this study, we integrated IMC into a clinical trial of hepatocellular carcinoma (HCC) patients undergoing combination therapy with PD-1 blockade and transarterial chemoembolization. We analyzed 281 regions of interest from 43 patients using a custom 40-marker IMC panel and developed a novel superpixel-based graph attention network, IMCSGAT, to model spatial cell interactions within the tumor microenvironment. IMCSGAT enabled accurate multitask prediction of key clinical features, including Barcelona Clinic Liver Cancer stage, trabecular histologic subtype, and treatment response. Compared to state-of-the-art methods, IMCSGAT achieved superior performance across all classification tasks. Spatial interaction analysis revealed that resident macrophage–centered interactions, particularly those with NK and T cells, were enriched in responders and predictive of therapeutic outcome. These findings were validated in a murine HCC model, reinforcing the role of innate immune architecture in shaping the treatment response. This study establishes IMCSGAT as a powerful spatial learning framework for high-dimensional IMC data, with potential applications in clinical outcome prediction and personalized therapy design for HCC. Our results provide a blueprint for the broader use of spatial analytics in precision oncology. The data underlying this article are available from the following link https://ngdc.cncb.ac.cn/omix/preview/Y37bGUet .

Unmanned aerial vehicles (UAVs)-assisted mobile crowdsensing (MCS) has emerged as a promising paradigm for data collection. However, challenges such as spectrum scarcity, device heterogeneity, and user mobility hinder efficient coordination of sensing, communication, and computation. To tackle these issues, we propose a joint optimization framework that integrates time slot partition for sensing, communication, and computation phases, resource allocation, and UAV 3D trajectory planning, aiming to maximize the amount of processed sensing data. The problem is formulated as a non-convex stochastic optimization and further modeled as a partially observable Markov decision process (POMDP) that can be solved by multi-agent deep reinforcement learning (MADRL) algorithm. To overcome the limitations of conventional multi-layer perceptron (MLP) networks, we design a novel MADRL algorithm with hybrid actor network. The newly developed method is based on heterogeneous agent proximal policy optimization (HAPPO), empowered by convolutional neural networks (CNN) for feature extraction and Kolmogorov-Arnold networks (KAN) to capture structured state-action dependencies. Extensive numerical results demonstrate that our proposed method achieves significant improvements in the amount of processed sensing data when compared with other benchmarks.

With smart buildings becoming increasingly reliant on cyber-physical systems to optimize energy efficiency and maintain occupant comfort, the growing risk of cyber-attacks poses significant threats to operational integrity. Despite the increasing deployment of advanced technologies in Building Automation Systems (BASs), a substantial gap remains in developing and deploying real-time mitigation strategies to defend against these vulnerabilities. This paper proposes a novel cyber-resilient control framework that integrates the Adaptive Model Predictive Control (AMPC) to enhance energy resilience of smart buildings. The framework aims to maintain acceptable levels of energy system performance under cyber-attacks by dynamically reconfiguring the control objectives and constraints of nominal MPC into AMPC based on cyber-attack detection outcomes to ensure system continuity. The feasibility and effectiveness of the proposed framework were demonstrated through a Hardware-in-the-Loop (HIL) experiment under a Denial-of-Service attack scenario, specifically a device reinitialization attack on a Variable Air Volume (VAV) terminal box. Results showed that the cyber-resilient control framework reduced temperature violations (i.e., unmet degree hours) by 76.3%, with a control mitigation response time within one minute, although it resulted in an 11.4% increase in power consumption. These findings underscore the potential of the proposed cyber-resilient framework to mitigate the impact of cyber-attacks, ensuring resilient operation and security for energy systems in smart buildings.

Vision Transformers (ViTs), extensively pre-trained on large-scale datasets, have become fundamental to foundation models, enabling adaptation to diverse downstream tasks. Existing adaptation methods typically require direct data access, rendering them infeasible in privacy-sensitive domains where clients are often reluctant to share their data. A straightforward solution may be sending the pre-trained ViT to clients for local adaptation, which poses issues of model intellectual property and incurs heavy client computation overhead. To address these issues, we propose a novel split adaptation (SA) method that enables effective downstream adaptation while protecting data and models. SA, inspired by split learning (SL), segments the pre-trained ViT into a frontend and a backend, with only the frontend shared with the client for data representation extraction. But unlike regular SL, SA replaces frontend parameters with low-bit quantized values, preventing direct exposure of the model. SA allows the client to add bi-level noise to the frontend and the extracted data representations, ensuring data protection. Accordingly, SA incorporates data-level and model-level out-of-distribution enhancements to mitigate noise injection’s impact. Our SA focuses on the challenging few-shot adaptation and adopts patch retrieval augmentation for overfitting alleviation. Extensive experiments on multiple datasets validate SA’s superiority over state-of-the-art methods and demonstrate its defense against advanced data reconstruction attacks while preventing model leakage with minimal computation cost on the client side. The source codes can be found at https://github.com/conditionWang/Split_Adaptation.

Recent advances in large language model (LLM) post-training have leveraged two distinct paradigms to enhance reasoning capabilities: reinforcement learning (RL) and knowledge distillation (KD). While RL enables the emergence of complex reasoning behaviors, it often suffers from low sample efficiency when the initial policy struggles to explore high-reward trajectories. Conversely, KD improves learning efficiency via mimicking the teacher model but tends to generalize poorly to out-of-domain scenarios. In this work, we present \textbf{KDRL}, a \textit{unified post-training framework} that jointly optimizes a reasoning model through teacher supervision (KD) and self-exploration (RL). Specifically, KDRL leverages policy gradient optimization to simultaneously minimize the reverse Kullback-Leibler divergence (RKL) between the student and teacher distributions while maximizing the expected rule-based rewards. We first formulate a unified objective that integrates GRPO and KD, and systematically explore how different KL approximations, KL coefficients, and reward-guided KD strategies affect the overall post-training dynamics and performance. Empirical results on multiple reasoning benchmarks demonstrate that KDRL outperforms GRPO and various KD baselines while achieving a favorable balance between performance and reasoning token efficiency. These findings indicate that integrating KD and RL serves as an effective and efficient strategy to train reasoning LLMs.

Unvoiced electromyography (EMG) is an effective communication tool for individuals unable to produce vocal speech. However, most prior methods rely on paired voiced and unvoiced EMG signals, along with speech data, for EMG-to-text conversion, which is not practical for such individuals. Given the rise of large language models (LLMs) in speech recognition, we explore their potential to understand unvoiced speech. To this end, we address the challenge of learning from unvoiced EMG alone and propose a novel EMG adaptor module that maps EMG features into an LLM's input space, achieving an average word error rate (WER) of 0.49 on a closed-vocabulary unvoiced EMG-to-text task. Even with a conservative data availability of just six minutes, our approach improves performance over specialized models by nearly 20%. While LLMs have been shown to be extendable to new language modalities -- such as audio -- understanding articulatory biosignals like unvoiced EMG remains more challenging. This work takes a crucial first step toward enabling LLMs to comprehend unvoiced speech using surface EMG.

Intelligent vehicles are systems tightly integrating computation, communication, and physical behavior. The recent proliferation of artificial intelligence, machine learning, the Internet of Things (IoT), and edge-fog–cloud computing envisions that intelligent vehicles are capable of innovative solutions to change our lifestyles. However, the potential benefits come along with new challenges and concerns on security and privacy. This special issue consists of 12 papers and covers broad research contributions, including 1) intrusion detection from in-vehicular networks to connected vehicles, drones, and global positioning systems; 2) authentication with matchmaking encryption, certificateless cryptography, and blockchains for Internet of Vehicles; 3) privacy protection with data sharing and cross-vehicle federated learning; and 4) secure data analysis supported by the cloud. The special issue seeks to assist theoretical analysis, system architecture design, emerging applications, and social impacts of intelligent vehicles.

Building heating, ventilation, and air conditioning (HVAC) systems account for nearly half of building energy consumption and [Formula: see text] of total energy consumption in the US. Their operation is also crucial for ensuring the physical and mental health of building occupants. Compared with traditional model-based HVAC control methods, the recent model-free deep reinforcement learning (DRL) based methods have shown good performance while do not require the development of detailed and costly physical models. However, these model-free DRL approaches often suffer from long training time to reach a good performance, which is a major obstacle for their practical deployment. In this work, we present a systematic approach to accelerate online reinforcement learning for HVAC control by taking full advantage of the knowledge from domain experts in various forms. Specifically, the algorithm stages include learning expert functions from existing abstract physical models and from historical data via offline reinforcement learning, integrating the expert functions with rule-based guidelines, conducting training guided by the integrated expert function and performing policy initialization from distilled expert function. Moreover, to ensure that the learned DRL-based HVAC controller can effectively keep room temperature within the comfortable range for occupants, we design a runtime shielding framework to reduce the temperature violation rate and incorporate the learned controller into it. Experimental results demonstrate up to 8.8X speedup in DRL training from our approach over previous methods, with low temperature violation rate.

Vision Transformers (ViTs), extensively pre-trained on large-scale datasets, have become essential to foundation models, allowing excellent performance on diverse downstream tasks with minimal adaptation. Consequently, there is growing interest in adapting pre-trained ViTs across various fields, including privacy-sensitive domains where clients are often reluctant to share their data. Existing adaptation methods typically require direct data access, rendering them infeasible under these constraints. A straightforward solution may be sending the pre-trained ViT to clients for local adaptation, which poses issues of model intellectual property protection and incurs heavy client computation overhead. To address these issues, we propose a novel split adaptation (SA) method that enables effective downstream adaptation while protecting data and models. SA, inspired by split learning (SL), segments the pre-trained ViT into a frontend and a backend, with only the frontend shared with the client for data representation extraction. But unlike regular SL, SA replaces frontend parameters with low-bit quantized values, preventing direct exposure of the model. SA allows the client to add bi-level noise to the frontend and the extracted data representations, ensuring data protection. Accordingly, SA incorporates data-level and model-level out-of-distribution enhancements to mitigate noise injection's impact on adaptation performance. Our SA focuses on the challenging few-shot adaptation and adopts patch retrieval augmentation for overfitting alleviation. Extensive experiments on multiple datasets validate SA's superiority over state-of-the-art methods and demonstrate its defense against advanced data reconstruction attacks while preventing model leakage with minimal computation cost on the client side. The source codes can be found at https://github.com/conditionWang/Split_Adaptation.

Objective: Evidence on long-term antibody persistence for the rabies vaccine administered using the four-dose Essen regimen is lacking. This study compared antibody persistence for the human diploid cell rabies vaccine (HDCV) administered using the four- versus five-dose Essen intramuscular regimen in post-exposure prophylaxis (PEP). Methods: This prospective cohort study enrolled patients vaccinated with the lyophilized HDCV for PEP who were grouped into four-dose and five-dose Essen groups. Rabies virus-neutralizing antibody (RVNA) detection was performed at 1 year or 3 years after initial vaccination. Results: In total, 180 and 184 patients were included in the four- and five-dose groups, respectively. The 1-year seroconversion (&gt;0.5 IU/mL) rates were similar in the five-dose and four-dose Essen groups (99.2% vs. 98.3%, p = 0.662), as were the 3-year seroconversion rates (98.4% vs. 98.3%, p &gt; 0.999). The median RVNA titer was significantly higher with the five-dose Essen regimen compared with the four-dose Essen regimen at 1 year (2.75 vs. 4.6 IU/mL, p = 0.002), and both groups had similar rates at 3 years (2.00 vs. 3.80 IU/mL, p = 0.443). Multivariable stepwise linear regression analysis showed that the five-dose Essen regimen was independently associated with higher serum RVNA titer compared to the four-dose Essen regimen (β = 0.175, p = 0.001), and 3 years after vaccination, was independently associated with a lower serum RVNA titer compared to 1 year (β = −1.06, p = 0.049). Conclusions: The four- and five-dose Essen regimens effectively produce durable immunogenicity, supporting the feasibility of implementing the four-dose Essen regimen for rabies immunization in China.

Background/purpose: Burning mouth syndrome (BMS) is a chronic condition characterized by intraoral burning sensation and orofacial pain but without oral mucosal lesions. The purpose of this study was to analyze the scientometric characteristics and research trends of BMS. Materials and methods: All the papers on BMS were comprehensively retrieved from the Scopus database. The years of publication were divided into before 2015 and 2015-2024 in the analysis of research trends. Results: index of 76. The related disorders of BMS were depression, xerostomia, pain, anxiety, glossodynia, taste disorder, nociception, paresthesia, analgesia, sleep disorder, diabetes mellitus, and neuralgia. Clonazepam was most common pharmacotherapy for BMS. After 2015, pharmacologic keywords including drug safety, aripiprazole, duloxetine, folic acid, hydrocortisone, and pregabalin were more frequent. Low level laser therapy, acupuncture, and cognitive behavioral therapy were the emerging nonpharmacologic strategies for BMS. Moreover, laboratory investigations on biomarkers, blood, genetics, interleukin 6, and tumor necrosis factor were more common. Various questionnaires, comorbidity, complication, anemia, hypertension, diabetes mellitus, and sleep were also more concerned after 2015. Conclusion: This scientometric study elucidated the current scenario and research trends of BMS, and would help in improving in reciprocal collaboration and communication for investigations on this condition.

Unvoiced electromyography (EMG) is an effective communication tool for individuals unable to produce vocal speech.However, most prior methods rely on paired voiced and unvoiced EMG signals, along with speech data, for unvoiced EMG-to-text conversion, which is not practical for these individuals.Given the rise of large language models (LLMs) in speech recognition, we explore their potential to understand unvoiced speech.To this end, we address the challenge of learning from unvoiced EMG alone and propose a novel EMG adaptor module that maps EMG features to an LLM's input space, achieving an average word error rate of 0.49 on a closed-vocabulary unvoiced EMG-to-text task.Even with a conservative data availability of just six minutes, our approach improves performance over specialized models by nearly 20%.While LLMs have been shown to be extendable to new language modalities-such as audio-understanding articulatory biosignals, like unvoiced EMG, is more challenging.This work takes a crucial first step toward enabling LLMs to comprehend unvoiced speech using surface EMG.

Although AR-HUD navigation has demonstrated potential in improving driving performance, most existing studies have focused on single-modality visual cues and have not sufficiently examined the integration of auditory and visual information, the influence of different visual presentation formats, or the optimal timing for delivering navigation prompts. These gaps are particularly evident in complex driving environments such as multi-branch intersections. This study addresses these issues by conducting a controlled driving simulation experiment to investigate the interactive effects of cue type and cue timing in a multimodal AR-HUD system. Two visual presentation formats, gradient navigation and boomerang-shaped icons, were synchronously paired with continuous auditory prompts and delivered at three timing conditions: 2000 m, 1500 m, and 1000 m before the intersection. Thirty participants completed driving tasks in a simulated multi-branch intersection scenario. Dependent variables included reaction time, eye-tracking metrics such as fixation count and average fixation duration, and subjective evaluations of situational awareness using the SART scale and cognitive load using the DALI scale. The results indicate that a prompt timing of 1500 m combined with gradient navigation achieved the highest situational awareness, the lowest cognitive load, the shortest reaction time, and the most efficient attention allocation. These findings provide empirical evidence for optimizing prompt design in multimodal AR-HUD systems, particularly in high-complexity intersections, and contribute to the refinement of human–machine interaction strategies to enhance driving safety.

Trajectory generation and trajectory prediction are two critical tasks in autonomous driving, which generate various trajectories for testing during development and predict the trajectories of surrounding vehicles during operation, respectively. In recent years, emerging data-driven deep learning-based methods have shown great promise for these two tasks in learning various traffic scenarios and improving average performance without assuming physical models. However, it remains a challenging problem for these methods to ensure that the generated/predicted trajectories are physically realistic. This challenge arises because learning-based approaches often function as opaque black boxes and do not adhere to physical laws. Conversely, existing model-based methods provide physically feasible results but are constrained by predefined model structures, limiting their capabilities to address complex scenarios. To address the limitations of these two types of approaches, we propose a new method that integrates kinematic knowledge into neural stochastic differential equations (SDE) and designs a variational autoencoder based on this latent kinematics-aware SDE (LK-SDE) to generate vehicle motions. Experimental results demonstrate that our method significantly outperforms both model-based and learning-based baselines in producing physically realistic and precisely controllable vehicle trajectories. Additionally, it performs well in predicting unobservable physical variables in the latent space.

Manufacturing workers face prolonged strenuous physical activities, impacting both financial aspects and their health due to work-related fatigue. Continuously monitoring physical fatigue and providing meaningful feedback is crucial to mitigating human and monetary losses in manufacturing workplaces. This study introduces a novel application of multimodal wearable sensors and machine learning techniques to quantify physical fatigue and tackle the challenges of real-time monitoring on the factory floor. Unlike past studies that view fatigue as a dichotomous variable, our central formulation revolves around the ability to predict multilevel fatigue, providing a more nuanced understanding of the subject's physical state. Our multimodal sensing framework is designed for continuous monitoring of vital signs, including heart rate, heart rate variability, skin temperature, and more, as well as locomotive signs by employing inertial motion units strategically placed at six locations on the upper body. This comprehensive sensor placement allows us to capture detailed data from both the torso and arms, surpassing the capabilities of single-point data collection methods. We developed an innovative asymmetric loss function for our machine learning model, which enhances prediction accuracy for numerical fatigue levels and supports real-time inference. We collected data on 43 subjects following an authentic manufacturing protocol and logged their self-reported fatigue. Based on the analysis, we provide insights into our multilevel fatigue monitoring system and discuss results from an in-the-wild evaluation of actual operators on the factory floor. This study demonstrates our system's practical applicability and contributes a valuable open-access database for future research.

Multi-agent patrolling is a key problem in a variety of domains such as intrusion detection, area surveillance, and policing, which involves repeated visits by a group of agents to specified points in an environment. While the problem is well-studied, most works do not provide performance guarantees and either do not consider agent attrition or impose significant communication requirements to enable adaptation. In this work, we present the Adaptive Heuristic-based Patrolling Algorithm, which is capable of adaptation to agent loss using minimal communication by taking advantage of Voronoi partitioning, and which meets guaranteed performance bounds. Additionally, we provide new centralized and distributed mathematical programming formulations of the patrolling problem, analyze the properties of Voronoi partitioning, and finally, show the value of our adaptive heuristic algorithm by comparison with various benchmark algorithms using physical robots and simulation based on the Robot Operating System (ROS) 2.

In this paper, we introduce a novel approach to enhancing neural network-based Simultaneous Localization and Mapping (SLAM) through the integration of model compression techniques and customized hardware architecture that focuses on micro-architectural and dataflow optimizations to improve computational efficiency and performance. Experiments across different scenarios demonstrate that the proposed approach achieves significant improvement.

Most existing speech disfluency detection techniques only rely upon acoustic data. In this work, we present a practical multimodal disfluency detection approach that leverages available video data together with audio. We curate an audiovisual dataset and propose a novel fusion technique with unified weight-sharing modality-agnostic encoders to learn the temporal and semantic context. Our resilient design accommodates real-world scenarios where the video modality may sometimes be missing during inference. We also present alternative fusion strategies when both modalities are assured to be complete. In experiments across five disfluency-detection tasks, our unified multimodal approach significantly outperforms Audio-only unimodal methods, yielding an average absolute improvement of 10% (i.e., 10 percentage point increase) when both video and audio modalities are always available, and 7% even when video modality is missing in half of the samples.

Large parallel applications have heavily used MPI (Massage Passing Interface) collectives that support portable and efficient group communication operations. MPI_Bcast is one of the most commonly used MPI collectives that broadcast data to all processes of the communication domain. However, traditional software-based broadcast algorithms fail to fully utilize modern interconnection networks’ advanced features such as offloading collectives to the network hardware for efficient group communications. Besides, the semantic gap between MPI_Bcast and hardware multicast of underlying interconnects presents challenges for offload-based algorithms to accelerate MPI_Bcast for a wide range of message sizes.In this paper, we propose a hardware-software co-design MPI_Bcast by efficiently leveraging the NIC-based collective offload provided by Tianhe-express interconnect, which completely precludes the involvement of CPU to accelerate message broadcast. We detail this broadcast mechanism that can be adaptively tuned to offload MPI_Bcast operations from the CPU to the NIC for various message and system sizes. In addition, we further propose a topology-aware broadcast design in conjunction with this offload method to significantly reduce the broadcast latency by constructing the optimal global inter-node communication tree. We implement and evaluate the proposed Tianhe-Express Offload-based Broadcast (TOB) design on Tianhe-2A and Tianhe-EP supercomputers. Extensive experiments have been conducted to evaluate TOB performance at both microbenchmark and application levels. Our solution offers up to 4.94x significant performance speedup at the microbenchmark level over state-of-the-art MPI libraries. For the application-level evaluation, our technique accelerates scientific applications by a maximum speedup of 1.34x.

Recently, the paradigm of pre-training and fine-tuning graph neural networks has been intensively studied and applied in a wide range of graph mining tasks. Its success is generally attributed to the structural consistency between pre-training and downstream datasets, which, however, does not hold in many real-world scenarios. Existing works have shown that the structural divergence between pre-training and downstream graphs significantly limits the transferability when using the vanilla fine-tuning strategy. This divergence leads to model overfitting on pre-training graphs and causes difficulties in capturing the structural properties of the downstream graphs. In this paper, we identify the fundamental cause of structural divergence as the discrepancy of generative patterns between the pre-training and downstream graphs. Furthermore, we propose G-Tuning to preserve the generative patterns of downstream graphs. Given a downstream graph G, the core idea is to tune the pre-trained GNN so that it can reconstruct the generative patterns of G, the graphon W. However, the exact reconstruction of a graphon is known to be computationally expensive. To overcome this challenge, we provide a theoretical analysis that establishes the existence of a set of alternative graphons called graphon bases for any given graphon. By utilizing a linear combination of these graphon bases, we can efficiently approximate W. This theoretical finding forms the basis of our model, as it enables effective learning of the graphon bases and their associated coefficients. Compared with existing algorithms, G-Tuning demonstrates consistent performance improvement in 7 in-domain and 7 out-of-domain transfer learning experiments.

Anomaly detection in time-series data is crucial for identifying faults, failures, threats, and outliers across a range of applications. Recently, deep learning techniques have been applied to this topic, but they often struggle in real-world scenarios that are complex and highly dynamic, e.g., the normal data may consist of multiple distributions, and various types of anomalies may differ from the normal data to different degrees. In this work, to tackle these challenges, we propose Distribution-Augmented Contrastive Reconstruction (DACR). DACR generates extra data disjoint from the normal data distribution to compress the normal data’s representation space, and enhances the feature extractor through contrastive learning to better capture the intrinsic semantics from time-series data. Furthermore, DACR employs an attention mechanism to model the semantic dependencies among multivariate time-series features, thereby achieving more robust reconstruction for anomaly detection. Extensive experiments conducted on nine benchmark datasets in various anomaly detection scenarios demonstrate the effectiveness of DACR in achieving new state-of-the-art time-series anomaly detection.

The growing interest in Large Language Models (LLMs) for specialized applications has revealed a significant challenge: when tailored to specific domains, LLMs tend to experience catastrophic forgetting, compromising their general capabilities and leading to a suboptimal user experience. Additionally, crafting a versatile model for multiple domains simultaneously often results in a decline in overall performance due to confusion between domains. In response to these issues, we present the RolE Prompting Guided Multi-Domain Adaptation (REGA) strategy. This novel approach effectively manages multi-domain LLM adaptation through three key components: 1) Self-Distillation constructs and replays general-domain exemplars to alleviate catastrophic forgetting. 2) Role Prompting assigns a central prompt to the general domain and a unique role prompt to each specific domain to minimize inter-domain confusion during training. 3) Role Integration reuses and integrates a small portion of domain-specific data to the general-domain data, which are trained under the guidance of the central prompt. The central prompt is used for a streamlined inference process, removing the necessity to switch prompts for different domains. Empirical results demonstrate that REGA effectively alleviates catastrophic forgetting and inter-domain confusion. This leads to improved domain-specific performance compared to standard fine-tuned models, while still preserving robust general capabilities.

Pattern recognition is a fundamental task in continuous sensing applications, but real-world scenarios often experience distribution shifts that necessitate learning generalizable representations for such tasks. This challenge is exacerbated with time-series data, which also exhibit inherent nonstationarity--variations in statistical and spectral properties over time. In this work, we offer a fresh perspective on learning generalizable representations for time-series classification by considering the phase information of a signal as an approximate proxy for nonstationarity and propose a phase-driven generalizable representation learning framework for time-series classification, PhASER. It consists of three key elements: 1) Hilbert transform-based augmentation, which diversifies nonstationarity while preserving task-specific discriminatory semantics, 2) separate magnitude-phase encoding, viewing time-varying magnitude and phase as independent modalities, and 3) phase-residual feature broadcasting, integrating 2D phase features with a residual connection to the 1D signal representation, providing inherent regularization to improve distribution-invariant learning. Extensive evaluations on five datasets from sleep-stage classification, human activity recognition, and gesture recognition against 13 state-of-the-art baseline methods demonstrate that PhASER consistently outperforms the best baselines by an average of 5% and up to 11% in some cases. Additionally, the principles of PhASER can be broadly applied to enhance the generalizability of existing time-series representation learning models.

The growing interest in Large Language Models (LLMs) for specialized applications has revealed a significant challenge: when tailored to specific domains, LLMs tend to experience catastrophic forgetting, compromising their general capabilities and leading to a suboptimal user experience.Additionally, crafting a versatile model for multiple domains simultaneously often results in a decline in overall performance due to confusion between domains.In response to these issues, we present the RolE Prompting Guided Multi-Domain Adaptation (REGA) strategy.This novel approach effectively manages multi-domain LLM adaptation through three key components: 1) Self-Distillation constructs and replays general-domain exemplars to alleviate catastrophic forgetting.2) Role Prompting assigns a central prompt to the general domain and a unique role prompt to each specific domain to minimize inter-domain confusion during training.3) Role Integration reuses and integrates a small portion of domainspecific data to the general-domain data, which are trained under the guidance of the central prompt.The central prompt is used for a streamlined inference process, removing the necessity to switch prompts for different domains.Empirical results demonstrate that REGA effectively alleviates catastrophic forgetting and inter-domain confusion.This leads to improved domain-specific performance compared to standard fine-tuned models, while still preserving robust general capabilities.

The recent development of cloud computing and edge computing shows great promise for the Connected and Automated Vehicle (CAV), by enabling CAVs to offload their massive on-board data and heavy computing tasks. Leveraging the Internet of Things (IoT) technology, different entities in the intelligent transportation system (e.g., vehicles, infrastructure, traffic management centers, etc.) get connected with each other, thus making the entire system smarter, faster, and more efficient. However, these advances also bring significant challenges to public authorities, industry, as well as scientific communities. In terms of system design and control, current cloud and edge architecture of CAVs need to be refined or even redesigned to better function under uncertainties in demand, and to better cooperate with existing conventional vehicles and infrastructure. From the performance assessment perspective, models and simulation tools based on artificial intelligence and big data have been widely developed for validation and evaluation of cloud computing and edge computing, but the validity of these models needs to be re-examined with field implementations. Finally, while the increasing connectivity among vehicles and infrastructures may help improve their perception of the environment and enable coordinated decision making, it also presents new challenges to ensure system safety and security, with inherent disturbances to wireless communication networks and also the inevitably larger attack surface that may be exploited by malicious attacks. In this tutorial, experts from academia and industry will introduce the trends and challenges of applying cloud and edge computing for CAVs, highlight representative works in the literature and discuss their limitations, present new promising solutions, and outline future directions in research and engineering. Particular focus will be given to methodologies and tools for building digital twin frameworks with cloud and edge computing for CAVs, quantitative and formal analysis for ensuring CAV safety under disturbances and uncertainties, system-level CAV security threat landscape and defense solution space, and experiences from practical deployment of cloud and edge computing for CAVs.

Autonomous Driving (AD) encounters significant safety hurdles in long-tail unforeseen driving scenarios, largely stemming from the non-interpretability and poor generalization of the deep neural networks within the AD system, particularly in out-of-distribution and uncertain data. To this end, this paper explores the integration of Large Language Models (LLMs) into AD systems, leveraging their robust common-sense knowledge and reasoning abilities. The proposed methodologies employ LLMs as intelligent decision-makers in behavioral planning, augmented with a safety verifier shield for contextual safety learning, for enhancing driving performance and safety. We present two key studies in a simulated environment: an adaptive LLM-conditioned Model Predictive Control (MPC) and an LLM-enabled interactive behavior planning scheme with a state machine. Demonstrating superior performance and safety metrics compared to state-of-the-art approaches, our approach shows the promising potential for using LLMs for autonomous vehicles.

MPI_Comm_split_type is a widely utilized operation in Message Passing Interface (MPI) programs, efficiently categorizing processes into two levels: inter-node and intra-node spaces. This division enhances the organization and management of communication, thereby reducing unnecessary communication overhead. Nevertheless, conventional implementation approaches suffer from two main deficiencies. Firstly, during node information collection, the hierarchy of processes remains indeterminate, resulting in significant redundant inter-node data transmission overhead. Secondly, the one-to-many communication pattern easily leads to communication bottlenecks, significantly increasing latency.To address these limitations, this paper introduces two optimization techniques aimed at minimizing the substantial redundancy in inter-node data transmission overhead. We integrate these techniques as dynamic libraries and evaluate them across two multi-core clusters, scaling experiments up to 512 computing nodes. Experimental results demonstrate that compared to state-of-the-art collective implementations, our optimization approach offers performance improvements ranging from 1.4 to 9.6 times. Importantly, these techniques exhibit outstanding scalability, maintaining exceptional performance in large-scale scenarios.

Objective: This study aims to understand the cognitive impact of latency in teleoperation and the related mitigation methods, using functional Near-Infrared Spectroscopy (fNIRS) to analyze functional connectivity. Background: Latency between command, execution, and feedback in teleoperation can impair performance and affect operators mental state. The neural underpinnings of these effects are not well understood. Method: A human subject experiment (n = 41) of a simulated remote robot manipulation task was performed. Three conditions were tested: no latency, with visual and haptic latency, with visual latency and no haptic latency. fNIRS and performance data were recorded and analyzed. Results: The presence of latency in teleoperation significantly increased functional connectivity within and between prefrontal and motor cortexes. Maintaining visual latency while providing real-time haptic feedback reduced the average functional connectivity in all cortical networks and showed a significantly different connectivity ratio within prefrontal and motor cortical networks. The performance results showed the worst performance in the all-delayed condition and best performance in no latency condition, which echoes the neural activity patterns. Conclusion: The study provides neurological evidence that latency in teleoperation increases cognitive load, anxiety, and challenges in motion planning and control. Real-time haptic feedback, however, positively influences neural pathways related to cognition, decision-making, and sensorimotor processes. Application: This research can inform the design of ergonomic teleoperation systems that mitigate the effects of latency.

Neural networks (NNs) playing the role of controllers have demonstrated impressive empirical performance on challenging control problems. However, the potential adoption of NN controllers in real-life applications has been significantly impeded by the growing concerns over the safety of these NN-controlled systems (NNCSs). In this work, we present POLAR-Express, an efficient and precise formal reachability analysis tool for verifying the safety of NNCSs. POLAR-Express uses Taylor model (TM) arithmetic to propagate TMs layer-by-layer across an NN to compute an overapproximation of the NN. It can be applied to analyze any feedforward NNs with continuous activation functions, such as ReLU, Sigmoid, and Tanh activation functions that cover the common benchmarks for NNCS reachability analysis. Compared with its earlier prototype POLAR, we develop a novel approach in POLAR-Express to propagate TMs more efficiently and precisely across ReLU activation functions, and provide parallel computation support for TM propagation, thus significantly improving the efficiency and scalability. Across the comparison with six other state-of-the-art tools on a diverse set of common benchmarks, POLAR-Express achieves the best verification efficiency and tightness in the reachable set analysis. POLAR-Express is publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/ChaoHuang2018/POLAR_Tool</uri> .

Smart buildings, especially Grid-interactive Efficient Buildings (GEBs), suffer from cyber-attacks and physical faults due to the integration of a large number of sensors and controls, connected devices, and associated communication networks. This study demonstrated a real-time advanced building resilient platform, called CYber Defense and REsilient System (CYDRES), which is deployable for existing and emerging Building Automation Systems (BASs). CYDRES aims to empower GEBs with cyber-attack-immune capabilities through multi-layer prevention and adaptation mechanisms to monitor, detect, and respond to cyber-attacks and physical operational faults. CYDRES is demonstrated through real-time experiments in a Hardware-in-the-Loop (HIL) testbed.