Although deep neural networks have significantly improved object detection accuracy, manually designing efficient architectures is costly, and NAS networks incur high computational overhead, limiting their applicability in resource-constrained or real-time scenarios. This study aims to develop an efficient neural architecture search method capable of rapidly generating low-latency, high-performance detection networks on edge devices. We propose Knowledge-Guided Neural Architecture Search (KGNAS). First, explicit knowledge from a hardware universal deployment framework is used to enrich the search space; second, implicit knowledge characterizes the impact of memory interactions in hierarchical operator blocks on latency; finally, a zero-cost proxy metric evaluates the feature extraction capability of the backbone without training. Experiments show that KGNAS can design efficient networks for various edge devices within one hour, reducing inference latency by 88% compared to DetNAS with only a 3% accuracy loss.

Personalized services are increasingly critical for autonomous systems such as smart vehicles, where Large Language Models (LLMs) enhance the driving experience. However, efficiently deploying and continuously updating these models is challenging due to resource constraints and dynamic environments. Existing approaches often neglect personalization and effective cross-edge collaboration, leading to inconsistent service quality and inefficient resource utilization. To address these gaps, we propose a vehicle-edge-cloud collaborative framework centered on our Intent-Driven Multi-Agent Communication (IDMAC) algorithm. IDMAC intelligently coordinates the LLM update life-cycle. It utilizes an intent encoder to model the short-term goals of each edge server from its trajectory. This explicit intent then guides attention mechanisms to facilitate efficient, goal-aligned communication, reducing network overhead and focusing on relevant information. This allows for the dynamic scheduling and routing of parameter-efficient model patches. In our architecture, lightweight models are deployed on vehicles, edge servers handle specialized fine-tuning, and the cloud periodically aggregates global knowledge to create unified updates. Experimental results show that the proposed framework significantly improves service quality and adaptability. IDMAC also strengthens inter-edge collaboration while reducing latency and energy consumption.

Network slicing has become a primary function of 5G/6G network resource management. However, the existing slicing schemes have not sufficiently discussed the reconfiguration optimization schemes brought by user behavior changes and mobile network environment fluctuations, leading to excessive service interruption rates and slice reconfiguration costs in dynamic environments. To address this problem, this paper proposes an End-to-end Steady-state Adaptive slicing method for Dynamic network state and load (ESAD). To realize the steady-state slicing decisions, ESAD takes the steady-state degree of network slicing and reconfiguration cost as the objective and constructs the slicing reconfiguration probability evaluation function based on the service load dynamics function and the time-varying function of the network channel conditions. To improve the predictability and steady-state degree of the slicing decision, ESAD introduces an ensemble deep learning method to predict the load service fluctuation based on the user behavior model and employs reinforcement learning to compute the channel dynamics boundary, which guides the slicing decision to balance the network dynamics factors. Experiments on quality of service assurance for 5G cloud game rendering class prove that ESAD can reduce reconfiguration probability and long-term reconfiguration cost by 49.45%-58.50% while improving system QoS assurance and capacity.

Implicit discourse relation recognition (IDRR) aims at recognizing the discourse relation between two text segments without an explicit connective. Recently, prompt learning has been applied to the IDRR task with great performance improvements over various neural network-based approaches. However, the discrete nature of the state-art-of-art prompting approach requires manual design of templates and answers, a big hurdle for its practical applications. In this paper, we propose a continuous version of prompt learning together with connective knowledge distillation, called AdaptPrompt, to reduce manual design efforts via continuous prompting while further improving performance via knowledge transfer. In particular, we design and train a few virtual tokens to form continuous templates and automatically select the most suitable one by gradient search in the embedding space. We also design an answer-relation mapping rule to generate a few virtual answers as the answer space. Furthermore, we notice the importance of annotated connectives in the training dataset and design a teacher-student architecture for knowledge transfer. Experiments on the up-to-date PDTB Corpus V3.0 validate our design objectives in terms of the better relation recognition performance over the state-of-the-art competitors.

The periodicity of time series has significantly advanced long-term forecasting and has attracted extensive research efforts. However, existing methods still suffer from neglecting critical low-energy periodic components and high sensitivity to outliers. To address these issues, we propose the PeriOdic Spectra Transition via MAmba Network (POSTMAN). This architecture introduces the periodic spectrum deviation forecasting (PSDF) technique, which extracts the shared spectrum to represent the common periodic features and generates deviation spectra to represent the specific periodic features. The shared periodic spectrum retains the critical low-amplitude components, while the deviation spectra preserve the slight differences between periods. To effectively leverage the differences, we develop a spectral convolution-enhanced Frequency Mamba Block (FMB), which learns the transition patterns of periodic deviation spectra and inhibits the impact of outliers during the transition procedure. Experiments on seven mainstream time series datasets demonstrate that POSTMAN outperforms existing state-of-the-art models in accuracy and robustness.

Early diagnosis and intervention of depression promote complete recovery, with its traditional clinical assessments depending on the diagnostic scales, clinical experience of doctors and patient cooperation. Recent researches indicate that functional near-infrared spectroscopy (fNIRS) based on deep learning provides a promising approach to depression diagnosis. However, collecting large fNIRS datasets within a standard experimental paradigm remains challenging, limiting the applications of deep networks that require more data. To address these challenges, in this paper, we propose an fNIRS-driven depression recognition architecture based on cross-modal data augmentation (fCMDA), which converts fNIRS data into pseudo-sequence activation images. The approach incorporates a time-domain augmentation mechanism, including time warping and time masking, to generate diverse data. Additionally, we design a stimulation task-driven data pseudo-sequence method to map fNIRS data into pseudo-sequence activation images, facilitating the extraction of spatial-temporal, contextual and dynamic characteristics. Ultimately, we construct a depression recognition model based on deep classification networks using the imbalance loss function. Extensive experiments are performed on the two-class depression diagnosis and five-class depression severity recognition, which reveal impressive results with accuracy of 0.905 and 0.889, respectively. The fCMDA architecture provides a novel solution for effective depression recognition with limited data.

Implicit Discourse Relation Recognition (IDRR) is an important task to classify the discourse relation sense between argument pairs without an explicit connective. Recently, prompt learning methods have demonstrated success in IDRR. However, prior work primarily transform IDRR into a connective-cloze task based on the masked language model (MLM), which limits the predicted connective to one single token. Also, they fail to fully exploit critical semantic features shared among various forms of templates. In this paper, we propose NCPrompt, an NSP-based prompt learning and Contrastive learning method for IDRR. Specifically, we transform the IDRR task into a next sentence prediction (NSP) task, which can allow various-length answer connectives and enlarge the construction of the verbalizer for prompt-learning methods. Also, we notice that various prompt templates naturally constitute positive samples applied for self-supervised contrastive learning. And the usage of NSP naturally creates hard negative samples by introducing different candidate connectives between the same example. To our knowledge, we are the first to combine self-supervised contrastive learning with prompt learning to obtain high-quality semantic representations. Experiments on the PDTB 3.0 corpus have demonstrated the effectiveness and superiority of our model.

In recent years, the operation and maintenance of Intelligent Computing Center(ICC) have been moving toward the direction of cloud-native, intelligent, and green. Applications deployed in ICC are increasingly adopting containerization technology and microservice architecture. Although the adoption of microservice architecture is expected to simplify the development, deployment, and maintenance of software, the resource competition of microservices can lead to the degradation of application performance and thus affect the quality of user experience. In addition, dependencies between microservices can introduce the cascading effect. These issues make it challenging for operators to achieve efficient resource management while maintaining the quality of user experience.In this paper, we analyze the impact of dependencies between microservices on application performance. We model the service dependency graph as a weighted graph and propose a hybrid resource allocation method based on Deep Q-network. With the goal of maximizing the application performance while minimizing resources usage, our method evaluates the importance of microservices based on the service dependency graph to set the reward function. Using a reinforcement learning algorithm, our method adjusts the compute and network resources allocated to an application based on dynamically changing available resources. Experimental results show that our method can converge quickly and improve the application performance by approximately 68% while reducing the use of compute and network resources.

The phenomenon of reflection superposition manifests itself as complex and prevalent in the real world, leading to the emergence of various simplified linear and non-linear formulas and related models. This study points out the limitations of existing models in scenarios where the light source is only on one side of the camera and conducts a theoretical, quantitative analysis of such scenarios. Based on this, we reconfigured the relevant parameters of the synthetic dataset used in training, compared them with the original settings, and contrasted them with the YTMT model. Experimental results show that the reconfigured parameters improve the model performance on our self-constructed dataset, especially in better understanding the high-level semantics of images.

Network applications involve massive heterogeneous data fusion and analysis. Artificial intelligence can significantly improve the convenience and user experience, but it requires a lot of storage, bandwidth, and computing resources. Multiaccess edge computing (MEC) extends intelligence services to IoT devices through offloading approaches and joint processing, which solves the resource bottleneck. However, designing advanced collaboration technology to offload tasks to MEC servers is still challenging. Heuristic algorithms and deep reinforcement learning (DRL)-based approaches have been proposed to offload tasks and minimize application latency. However, heuristic algorithms heavily depend on accurate mathematical models for the MEC system, and DRL does not make fair use of the relationship between devices in the MEC graph. To solve this, we propose a task offloading mechanism based on graph neural network (GNN), which can directly learn on graph data with messages passing and aggregation. We propose a graph reinforcement learning-based offloading (GRLO) framework, which models MEC as an acyclic graph and the offloading policy by graph state migration. GRLO combines GNN with the actor-critic network and trains offloading decision makers without labels. To efficiently train the GRLO, we propose a method that quickly explores action space and approaches the optimal solution. The numerical results show that the GRLO has lower latency compared to baselines while having generalization ability to new environments and topologies. Moreover, we verified the effectiveness of GRLO on a prototype.

In upcoming sixth generation (6G) networks, it is a critical challenge to support a plethora of innovative services across wide-area networks. To realize the dedicated QoS provisioning and meet the diverse quality of service (QoS) requirements of services in terms of criteria like bandwidth, delay, jitter and loss ratio, we proposed an AI-driven on-demand routing framework to support the diverse end-to-end QoS Provisioning in large-scale wide-area networks. Specifically, we make further efforts on solving the instability and non-convergence issues of the AI-driven routing algorithm and enhance it with the assistance of expert knowledge on traffic engineering and the latest advance on reinforcement learning. Furthermore, the simulation results show that our algorithm outperforms other benchmark routing algorithms with efficient learning and a significant reduction in delay, jitter and loss ratio by the traffic data sets of the real-world wide-area networks.

Chinese event detection (CED) often suffers from the problem of word-trigger mismatches due to the absence of natural delimiters in Chinese sentences. Recently many neural network models have been proposed to solve this problem by learning both character-level and word-level representation, however, their performance is limited by the insufficient ability of capturing long-range dependencies. In this paper, we propose a Mesh Transformer to model character–word interaction for the CED task. The Mesh Transformer has three kinds of nodes: character nodes, word nodes and a relay node. Compared with the fully-connected standard Transformer, each character node interacts with its neighboring character nodes and its self-matched word nodes to capture local semantic information in the Mesh Transformer. Additionally, the relay node is connected to each character node and word node, where long-range dependencies can be captured by a two-step update. Furthermore, entity knowledge is utilized to deal with the label confusion problem by improving the capability of modeling semantic representation. Experiment results on the ACE 2005 Chinese dataset show that the proposed model achieves superior performance than the peer and state-of-the-art ones in terms of F1-score on both trigger identification and trigger classification.

In recent years, research on predicting future events using temporal knowledge graphs by leveraging their rich structural and historical information has just begun. Due to their interesting application potential, they have gained more attention from researchers. Existing methods mainly focus on using evolution representation learning to aggregate and propagate recent graph structural information or integrate historical memory to infer the future. However, simply considering the historical repeated facts during reasoning is insufficient. Therefore, we propose a novel model called LogE-Net (Logic Evolution network) that simultaneously considers short-term facts and logical rules. Additionally, we propose a mutual-information control mechanism based on contrastive learning to control the freedom of temporal evolution. Finally, experiments on two public datasets demonstrate that our model consistently outperforms the state-of-the-art baseline on entity prediction and relation prediction tasks.

Text-based video retrieval is a well-studied task aimed at retrieving relevant videos from a large collection in response to a given text query. Most existing TVR works assume that videos are already trimmed and fully relevant to the query thus ignoring that most videos in real-world scenarios are untrimmed and contain massive irrelevant video content. Moreover, as users' queries are only relevant to video events rather than complete videos, it is also more practical to provide specific video events rather than an untrimmed video list. In this paper, we introduce a challenging but more realistic task called Multi-Granularity Video Content Retrieval (MGVCR), which involves retrieving both video files and specific video content with their temporal locations. This task presents significant challenges since it requires identifying and ranking the partial relevance between long videos and text queries under the lack of temporal alignment supervision between the query and relevant moments. To this end, we propose a novel unified framework, termed, Joint Searching and Grounding (JSG). It consists of two branches: (1) a glance branch that coarsely aligns the query and moment proposals using inter-video contrastive learning, and (2) a gaze branch that finely aligns two modalities using both inter- and intra-video contrastive learning. Based on the glance-to-gaze design, our JSG method learns two separate joint embedding spaces for moments and text queries using a hybrid synergistic contrastive learning strategy. Extensive experiments on three public benchmarks, i.e., Charades-STA, DiDeMo, and ActivityNet-Captions demonstrate the superior performance of our JSG method on both video-level retrieval and event-level retrieval subtasks. Our open-source implementation code is available at https://github.com/CFM-MSG/Code_JSG.

Text-based person retrieval aims at searching for a pedestrian image from multiple candidates with textual descriptions. It is challenging due to uncertain cross-modal alignments caused by the large intra-class variations. To address the challenge, most existing approaches rely on various attention mechanisms and auxiliary information, yet still struggle with the uncertain cross-modal alignments arising from significant intra-class variation, leading to coarse retrieval results. To this end, we propose a novel framework termed Deep Cross-modal Evidential Learning (DCEL), which deploys evidential deep learning to consider the cross-modal alignment uncertainty. Our DCEL model comprises three components: (1) Bidirectional Evidential Learning, which models alignment uncertainty to measure and mitigate the influence of large intra-class variation; (2) Multi-level Semantic Alignment, which leverages a proposed Semantic Filtration module and image-text similarity distribution to facilitate cross-modal alignments; (3) Cross-modal Relation Learning, which reasons about latent correspondences between multi-level tokens of image and text. Finally, we integrate the advantages of the three proposed components to enhance the model to achieve reliable cross-modal alignments. Our DCEL method consistently outperforms more than ten state-of-the-art methods in supervised, weakly supervised, and domain generalization settings on three benchmarks: CUHK-PEDES, ICFG-PEDES, and RSTPReid.

Artificial intelligence has been adopted to facilitate monitoring, operation, and decision in the logistics field.Logistics robots with environment perception capability have been used to improve warehousing efficiency in logistics systems.However, autonomous mobile robots face computationally intensive and real-time demanding tasks such as navigation, localization, and obstacle avoidance.In this article, we present EventTube, an edge computing based event-aware system that can efficiently discover events from the video data captured by RGB-Monoculars and collaborate with individual devices to make timely decisions.EventTube deploys a semantic context extraction pipeline on edge servers to aggregate video streams from mobile robots and feed a few keyframes, including the start and end of the specific events to the successive perception pods, accelerating logistics robots' response speed.The event-related model parameters are trained and updated online on a server.The video data collected at the warehouse site for our mobile robots show that EventTube significantly improves parcel delivery efficiency without affecting regular deliveries.

Prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition (IDRR). However, diverse relation senses may benefit from different attention mechanisms. We also argue that some linguistic relation in between two words can be further exploited for IDRR. This paper proposes a Multi-Attentive Neural Fusion (MANF) model to encode and fuse both semantic connection and linguistic evidence for IDRR. In MANF, we design a Dual Attention Network (DAN) to learn and fuse two kinds of attentive representation for arguments as its semantic connection. We also propose an Offset Matrix Network (OMN) to encode the linguistic relations of word-pairs as linguistic evidence. Our MANF model achieves the state-of-the-art results on the PDTB 3.0 corpus.

In human activity recognition, it is important to first recognize the position of a user. Although numerous fingerprinting schemes have been researched for indoor localization, they work in a stand-alone mode without considering the potentials of exploiting multiple requesting users’ online fingerprints. In this paper, we propose an emerging online collaborative localization paradigm to serve the scenarios of multiple users requesting localizations at the same time, which enables a kind of collective calibration by exploiting latent relations in between requesting users in both the signal space and physical space. The proposed scheme is called MOCLoc, which first applies the multidimensional scaling (MDS) technique to compute online users’ virtual locations based on their online fingerprints. A new virtual location credibility is next computed for each selected reference point (RP) based on the similarity between its virtual distances and physical distances to other online users. We also compute a collaborative credibility and signal credibility and fuse all three kinds of credibilities to compute a new RP ranking weight that is used to output the user final location. Furthermore, we propose to include a device calibration module based on the mean-subtraction fingerprint transformation to differentiate the collaborative calibration process for heterogeneous online devices. Experiments on field measurements validate the effectiveness of the proposed MOCLoc scheme in terms of further reduced localization errors.

Event argument extraction aims at identifying event arguments from texts as well as determining their respective roles in an event. Despite some neural networks applied for this task, their performance are still not satisfactory due to the following shortcomings. Syntactic information were not well explored; Event arguments were independently extracted; Pattern knowledge were not explicitly exploited. In this paper, we propose a Syntactic Distance Sensitive Neural Network model to tackle these problems. Our model first captures long-range dependencies in between event triggers and event arguments through performing graph convolution over syntactic trees, where we introduce syntactic distance to weight the importance of each word. Furthermore, we design an argument interaction module to mine argument-argument interactions according to the shortest dependency distances in between arguments. To enjoy pattern knowledge, we design a pattern-aware argument classification module to ensure the reasonability of extracted arguments. Extensive experiments have validated the superiority of the proposed model, which achieves the state-of-the-art results in terms of better F1-score on both argument identification and role classification.