Cong Huy Nguyen, Son Dinh Nguyen, Guanlin Li, Tuan Dung Nguyen, Aditya Narayan Sankaran, Mai Huy Thong, Thanh Trung Nguyen, Mai Hong Son, Reza Farahbakhsh, Phi Le Nguyen, Noel Crespi. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Attenuation artifacts remain a significant challenge in cardiac Myocardial Perfusion Imaging (MPI) using Single-Photon Emission Computed Tomography (SPECT), often compromising diagnostic accuracy and reducing clinical interpretability. While hybrid SPECT/CT systems mitigate these artifacts through CT-derived attenuation maps, their high cost, limited accessibility, and added radiation exposure hinder widespread clinical adoption. In this study, we propose a novel CT-free solution to attenuation correction in cardiac SPECT. Specifically, we introduce Physics-aware Attenuation Correction Diffusion Model (PADM), a diffusion-based generative method that incorporates explicit physics priors via a teacher–student distillation mechanism. This approach enables attenuation artifact correction using only Non-Attenuation-Corrected (NAC) input, while still benefiting from physics-informed supervision during training. To support this work, we also introduce CardiAC, a comprehensive dataset comprising 424 patient studies with paired NAC and Attenuation-Corrected (AC) reconstructions, alongside high-resolution CT-based attenuation maps. Extensive experiments demonstrate that PADM outperforms state-of-the-art generative models, delivering superior reconstruction fidelity across both quantitative metrics and visual assessment.

Attenuation artifacts remain a critical challenge in cardiac Myocardial Perfusion Imaging (MPI) using Single-Photon Emission Computed Tomography (SPECT), often degrading diagnostic accuracy and clinical interpretability. While hybrid SPECT and Computed Tomography (CT) systems mitigate these artifacts using CT-derived attenuation maps, their high cost, radiation exposure, and limited accessibility restrict widespread clinical use. To address these challenges, we propose DiffCAS, an inference-time CT-free diffusion model for physics-aware multi-slice attenuation correction in cardiac SPECT. DiffCAS integrates a Brownian Bridge diffusion process with physics-guided supervision, enabling the generation of attenuation-corrected (AC) images directly from non-attenuation-corrected (NAC) inputs. Specifically, a physics-aware reconstruction module predicts voxel-wise attenuation coefficients and path lengths, then combines them via the Beer-Lambert law into an attenuation correction factor applied at each diffusion step, keeping the AC images physically consistent. The model introduces two key innovations that jointly enhance structural understanding and physics consistency. The first is multi-slice contextual learning, which captures cross-slice anatomical dependencies and improves spatial coherence in reconstructed images. The second is the 3D Computed Tomography Vision Transformer that models long-range volumetric structures and provides physics-consistent attenuation priors to guide the diffusion process. To enable CT-free attenuation correction, DiffCAS introduces the teacher-student distillation framework that transfers physics-informed knowledge from CT-conditioned training to a student network that requires no CT input at inference time, ensuring stability and interpretability. Evaluations on the CardiAC dataset, which comprises 424 patient studies with paired NAC and AC, and CT-based attenuation maps, demonstrate the strong performance of DiffCAS, evaluated using global pixel-level metrics and myocardium-specific clinical metrics. The proposed method surpasses state-of-the-art image generative methods, achieving superior reconstruction accuracy, structural consistency, and diagnostic reliability. These results highlight the proposed DiffCAS as a clinically promising, inference-time CT-free solution for attenuation correction in cardiac SPECT imaging.

The rapid advancement of artificial intelligence (AI) systems such as large language models (LLMs) and multimodal learning frameworks has transformed digital content creation and manipulation. Traditional visual editing tools require significant expertise, limiting accessibility. Recent strides in instruction-guided editing have enabled intuitive interaction with visual content, using natural language as a bridge between user intent and complex editing operations. This survey reviews how AI-implemented instruction-guided image editing models – including approaches rooted in generative adversarial networks and diffusion models – empower users to achieve precise visual modifications without deep technical knowledge. By synthesizing over 100 publications, we examine multimodal integration for fine-grained content control, compare existing literature, and highlight how these AI applications support creative visual storytelling, design workflows, and multimedia production. We also identify key challenges to stimulate further research. Interested readers are encouraged to access our repository at https://github.com/tamlhp/awesome-instructional-editing .

Dynamic Time Warping (DTW) is a widely used metric for time series matching. However, when applied to few-shot action recognition (FSAR), DTW often encounters the "identical matching" issue, where multiple frames from one video are matched to a single frame from another. To address this, we introduce FTP-FSAR, a novel metric-based FSAR approach designed to mitigate this challenge. FTP-FSAR proposes an innovative alignment metric that incorporates temporal priors, guiding the matching process by encouraging the alignment of frames with similar temporal progression, thus improving the accuracy of frame matching. Additionally, FTP-FSAR integrates a dual framework, combining a foundation model with transductive learning to optimize feature extraction. Extensive experiments across multiple datasets demonstrate that FTP-FSAR outperforms existing methods, achieving the best results in 3 out of 4 benchmarks across 1-shot, 3-shot, and 5-shot settings, with performance improvements of up to 4.5%.

Visual grounding (VG) is the capability to identify the specific regions in an image associated with a particular text description. In medical imaging, VG enhances interpretability by highlighting relevant pathological features corresponding to textual descriptions, improving model transparency and trustworthiness for wider adoption of deep learning models in clinical practice. Current models struggle to associate textual descriptions with disease regions due to inefficient attention mechanisms and a lack of fine-grained token representations. In this paper, we empirically demonstrate two key observations. First, current VLMs assign high norms to background tokens, diverting the model's attention from regions of disease. Second, the global tokens used for cross-modal learning are not representative of local disease tokens. This hampers identifying correlations between the text and disease tokens. To address this, we introduce simple, yet effective Disease-Aware Prompting (DAP) process, which uses the explainability map of a VLM to identify the appropriate image features. This simple strategy amplifies disease-relevant regions while suppressing background interference. Without any additional pixel-level annotations, DAP improves visual grounding accuracy by 20.74% compared to state-of-the-art methods across three major chest X-ray datasets.

Deep neural networks often suffer performance drops when test data distribution differs from training data. Domain Generalization (DG) aims to address this by focusing on domain-invariant features or augmenting data for greater diversity. However, these methods often struggle with limited training domains or significant gaps between seen (training) and unseen (test) domains. To enhance DG robustness, we hypothesize that it is essential for the model to be trained on data from domains that closely resemble unseen test domains-an inherently difficult task due to the absence of prior knowledge about the unseen domains. Accordingly, we propose ConstStyle, a novel approach that leverages a unified domain to capture domain-invariant features and bridge the domain gap with theoretical analysis. During training, all samples are mapped onto this unified domain, optimized for seen domains. During testing, unseen domain samples are projected similarly before predictions. By aligning both training and testing data within this unified domain, ConstStyle effectively reduces the impact of domain shifts, even with large domain gaps or few seen domains. Extensive experiments demonstrate that ConstStyle consistently outperforms existing methods across diverse scenarios. Notably, when only a limited number of seen domains are available, ConstStyle can boost accuracy up to 19.82\% compared to the next best approach.

Understanding the inner workings of neural networks is essential for enhancing model performance and interpretability. Current research predominantly focuses on examining the connection between individual neurons and the model's final predictions, which suffers from challenges in interpreting the internal workings of the model, particularly when neurons encode multiple unrelated features. In this paper, we propose a novel framework that transitions the focus from analyzing individual neurons to investigating groups of neurons, shifting the emphasis from neuron-output relationships to the functional interactions between neurons. Our automated framework, NeurFlow, first identifies core neurons and clusters them into groups based on shared functional relationships, enabling a more coherent and interpretable view of the network’s internal processes. This approach facilitates the construction of a hierarchical circuit representing neuron interactions across layers, thus improving interpretability while reducing computational costs. Our extensive empirical studies validate the fidelity of our proposed NeurFlow. Additionally, we showcase its utility in practical applications such as image debugging and automatic concept labeling, thereby highlighting its potential to advance the field of neural network explainability.

Medical Large Multi-modal Models (LMMs) have demonstrated remarkable capabilities in medical data interpretation. However, these models frequently generate hallucinations contradicting source evidence, particularly due to inadequate localization reasoning. This work reveals a critical limitation in current medical LMMs: instead of analyzing relevant pathological regions, they often rely on linguistic patterns or attend to irrelevant image areas when responding to disease-related queries. To address this, we introduce HEAL-MedVQA (Hallucination Evaluation via Localization MedVQA), a comprehensive benchmark designed to evaluate LMMs' localization abilities and hallucination robustness. HEAL-MedVQA features (i) two innovative evaluation protocols to assess visual and textual shortcut learning, and (ii) a dataset of 67K VQA pairs, with doctor-annotated anatomical segmentation masks for pathological regions. To improve visual reasoning, we propose the Localize-before-Answer (LobA) framework, which trains LMMs to localize target regions of interest and self-prompt to emphasize segmented pathological areas, generating grounded and reliable answers. Experimental results demonstrate that our approach significantly outperforms state-of-the-art biomedical LMMs on the challenging HEAL-MedVQA benchmark, advancing robustness in medical VQA.

This paper studies the black-box optimization task which aims to find the maxima of a black-box function using a static set of its observed input-output pairs. This is often achieved via learning and optimizing a surrogate function with that offline data. Alternatively, it can also be framed as an inverse modeling task that maps a desired performance to potential input candidates that achieve it. Both approaches are constrained by the limited amount of offline data. To mitigate this limitation, we introduce a new perspective that casts offline optimization as a distributional translation task. This is formulated as learning a probabilistic bridge transforming an implicit distribution of low-value inputs (i.e., offline data) into another distribution of high-value inputs (i.e., solution candidates). Such probabilistic bridge can be learned using low- and high-value inputs sampled from synthetic functions that resemble the target function. These synthetic functions are constructed as the mean posterior of multiple Gaussian processes fitted with different parameterizations on the offline data, alleviating the data bottleneck. The proposed approach is evaluated on an extensive benchmark comprising most recent methods, demonstrating significant improvement and establishing a new state-of-the-art performance. Our code is publicly available at https://github.com/cuong-dm/ROOT.

Multimodal federated learning in real-world settings often encounters incomplete and heterogeneous data across clients. This results in misaligned local feature representations that limit the effectiveness of model aggregation. Unlike prior work that assumes either differing modality sets without missing input features or a shared modality set with missing features across clients, we consider a more general and realistic setting where each client observes a different subset of modalities and might also have missing input features within each modality. To address the resulting misalignment in learned representations, we propose a new federated learning framework featuring locally adaptive representations based on learnable client-side embedding controls that encode each client’s data-missing patterns. These embeddings serve as reconfiguration signals that align the globally aggregated representation with each client's local context, enabling more effective use of shared information. Furthermore, the embedding controls can be algorithmically aggregated across clients with similar data-missing patterns to enhance the robustness of reconfiguration signals in adapting the global representation. Empirical results on multiple federated multimodal benchmarks with diverse data-missing patterns across clients demonstrate the efficacy of the proposed method, achieving up to 36.45\% performance improvement under severe data incompleteness. The method is also supported by a theoretical analysis with an explicit performance bound that matches our empirical observations. Our source codes are provided at [https://github.com/nmduonggg/PEPSY](https://github.com/nmduonggg/PEPSY)

Vision-Language Foundation Models (VLMs), trained on large-scale multimodal datasets, have driven significant advances in Artificial Intelligence (AI) by enabling rich cross-modal reasoning. Despite their success in general domains, applying these models to medical imaging remains challenging due to the limited availability of diverse imaging modalities and multilingual clinical data. Most existing medical VLMs are trained on a subset of imaging modalities and focus primarily on high-resource languages, thus limiting their generalizability and clinical utility. To address these limitations, we introduce a novel Vietnamese-language multimodal medical dataset consisting of 2,757 whole-body PET/CT volumes from independent patients and their corresponding full-length clinical reports. This dataset is designed to fill two pressing gaps in medical AI development: (1) the lack of PET/CT imaging data in existing VLMs training corpora, which hinders the development of models capable of handling functional imaging tasks; and (2) the underrepresentation of low-resource languages, particularly the Vietnamese language, in medical vision-language research. To the best of our knowledge, this is the first dataset to provide comprehensive PET/CT-report pairs in Vietnamese. We further introduce a training framework to enhance VLMs' learning, including data augmentation and expert-validated test sets. We conduct comprehensive experiments benchmarking state-of-the-art VLMs on downstream tasks, including medical report generation and visual question answering. The experimental results show that incorporating our dataset significantly improves the performance of existing VLMs. However, despite these advancements, the models still underperform on clinically critical criteria, particularly the diagnosis of lung cancer, indicating substantial room for future improvement. We believe this dataset and benchmark will serve as a pivotal step in advancing the development of more robust VLMs for medical imaging, particularly in low-resource languages, and improving their clinical relevance in Vietnamese healthcare.

Vision-based sign language recognition is an extensively researched problem aimed at advancing communication be-tween deaf and hearing individuals. Numerous Sign Lan-guage Recognition (SLR) datasets have been introduced to promote research in this field, spanning multiple languages, vocabulary sizes, and signers. However, most existing pop-ular datasets focus predominantly on the frontal view of signers, neglecting visual information from other perspec-tives. In practice, many sign languages contain words that have similar hand movements and expressions, making it challenging to differentiate between them from a single frontal view. Although a few studies have proposed sign language datasets using multi-view data, these datasets remain limited in vocabulary size and scale, hindering their gener-alizability and practicality. To address this issue, we in-troduce a new large-scale, multi-view sign language recog-nition dataset spanning 1,000 glosses and 30 signers, re-sulting in over 84,000 multi-view videos. To the best of our knowledge, this is the first multi-view sign language recognition dataset of this scale. In conjunction with of-fering a comprehensive dataset, we perform extensive ex-periments to assess the performance of state-of-the-art Sign Language Recognition models utilizing on our dataset. The findings indicate that utilizing multi-view data substantially enhances model accuracy across all models, with a maxi-mum performance improvement of up to 19.75% compared to models trained on single-view data. Our dataset and baseline models are publicly accessible on GitHub11Available at https://github.com/Etdihatthoc/Multi-VSL_WACV_2025.

Positron Emission Tomography (PET) and Computed Tomography (CT) are essential for diagnosing, staging, and monitoring various diseases, particularly cancer. Despite their importance, the use of PET/CT systems is limited by the necessity for radioactive materials, the scarcity of PET scanners, and the high cost associated with PET imaging. In contrast, CT scanners are more widely available and significantly less expensive. In response to these challenges, our study addresses the issue of generating PET images from CT images, aiming to reduce both the medical examination cost and the associated health risks for patients. Our contributions are twofold: First, we introduce a conditional diffusion model named CPDM, which, to our knowledge, is one of the initial attempts to employ a diffusion model for translating from CT to PET images. Second, we provide the largest CT-PET dataset to date, comprising 2,028,628 paired CT-PET images, which facilitates the training and evaluation of CT-to-PET translation models. For the CPDM model, we incorporate domain knowledge to develop two conditional maps: the Attention map and the Attenuation map. The former helps the diffusion process focus on areas of interest, while the latter improves PET data correction and ensures accurate diagnostic information. Experimental evaluations across various benchmarks demonstrate that CPDM surpasses existing methods in generating high-quality PET images in terms of multiple metrics. The source code and data samples are available at https://github.com/thanhhff/CPDM.

Abstract The evolution of edge computing has advanced the accessibility of E-health recommendation services, encompassing areas such as medical consultations, prescription guidance, and diagnostic assessments. Traditional methodologies predominantly utilize centralized recommendations, relying on servers to store client data and dispatch advice to users. However, these conventional approaches raise significant concerns regarding data privacy and often result in computational inefficiencies. E-health recommendation services, distinct from other recommendation domains, demand not only precise and swift analyses but also a stringent adherence to privacy safeguards, given the users’ reluctance to disclose their identities or health information. In response to these challenges, we explore a new paradigm called on-device recommendation tailored to E-health diagnostics, where diagnostic support (such as biomedical image diagnostics), is computed at the client level. We leverage the advances of federated learning to deploy deep learning models capable of delivering expert-level diagnostic suggestions on clients. However, existing federated learning frameworks often deploy a singular model across all edge devices, overlooking their heterogeneous computational capabilities. In this work, we propose an adaptive federated learning framework utilizing BlockNets, a modular design rooted in the layers of deep neural networks, for diagnostic recommendation across heterogeneous devices. Our framework offers the flexibility for users to adjust local model configurations according to their device’s computational power. To further handle the capacity skewness of edge devices, we develop a data-free knowledge distillation mechanism to ensure synchronized parameters of local models with the global model, enhancing the overall accuracy. Through comprehensive experiments across five real-world datasets, against six baseline models, within six experimental setups, and various data distribution scenarios, our architecture demonstrates unparalleled performance and robustness in terms of both accuracy and efficiency.

Abstract As the adoption of explainable AI (XAI) continues to expand, the urgency to address its privacy implications intensifies. Despite a growing corpus of research in AI privacy and explainability, there is little attention on privacy-preserving model explanations. This article presents the first thorough survey about privacy attacks on model explanations and their countermeasures. Our contribution to this field comprises a thorough analysis of research papers with a connected taxonomy that facilitates the categorization of privacy attacks and countermeasures based on the targeted explanations. This work also includes an initial investigation into the causes of privacy leaks. Finally, we discuss unresolved issues and prospective research directions uncovered in our analysis. This survey aims to be a valuable resource for the research community and offers clear insights for those new to this domain. To support ongoing research, we have established an online resource repository, which will be continuously updated with new and relevant findings.

Federated Learning (FL) has emerged as an effective paradigm allowing multiple parties to collaboratively train a global model while protecting their private data. However, it is observed that the performance of FL approaches tends to degrade significantly when data are sparsely distributed across clients with small datasets. This is referred to as the sparse-and-scarce challenge, where data held by each client is both sparse (does not contain examples to all classes) and scarce (small dataset). Sparse-and-scarce data diminishes the generalizability of clients’ data, leading to intensive over-fitting and massive domain shifts in the local models and, ultimately, decreasing the aggregated model's performance. Interestingly, while this scenario is a specific manifestation of the well-known non-IID11This refers to the generic situation where local data distributions are not identical and independently distributed.challenge in FL, it has not been distinctly addressed. Our empirical investigation highlights that generic approaches to the non-IID challenge often prove inadequate in mitigating the sparse-and-scarce issue. To bridge this gap, we develop SAFA, a novel FL algorithm that specifically addresses the sparse-and-scarce challenge via a novel continual model iteration procedure. SAFA maximally exposes local models to the inter-client diversity of data with minimal effects of catastrophic forgetting. Our experiments show that SAFA outperforms existing FL solutions, up to 17.86%, compared to the prominent baseline. The code is accessible viahttps://github.com/HungNguyen20/SAFA.

Today, computer systems hold large amounts of personal data. Yet while such an abundance of data allows breakthroughs in AI, and especially machine learning, its existence can be a threat to user privacy, and it can weaken the bonds of trust between humans and AI. Recent regulations now require that, on request, private information about a user must be removed both from computer systems and from machine learning models—this legislation is more colloquially called “the right to be forgotten.” While removing data from back-end databases should be straightforward, it is not sufficient in the AI context as machine learning models often “remember” the old data. Contemporary adversarial attacks on trained models have proven that we can learn whether an instance or an attribute belonged to the training data. This phenomenon calls for a new paradigm, namely machine unlearning , to make machine learning models forget about particular data. It turns out that recent works on machine unlearning have not been able to completely solve the problem due to the lack of common frameworks and resources. Therefore, this article aspires to present a comprehensive examination of machine unlearning’s concepts, designs, methods, and applications. Specifically, as a category collection of cutting-edge studies, the intention behind this article is to serve as a comprehensive resource for researchers and practitioners seeking an introduction to machine unlearning and its formulations, design criteria, removal requests, algorithms, and applications. In addition, we aim to highlight the key findings, current trends, and new research areas that have not yet featured the use of machine unlearning but could benefit greatly from it. We hope that this survey serves as a valuable resource for machine learning researchers and those seeking to innovate privacy technologies. Our resources are publicly available at https://github.com/tamlhp/awesome-machine-unlearning .

Recommendation systems play a crucial role in providing web-based suggestion utilities by leveraging user behavior, preferences, and interests. In the context of privacy concerns and the proliferation of handheld devices, federated recommender systems have emerged as a promising solution. These systems allow each client to train a local model and exchange only the model updates with a central server, thus preserving data privacy. However, certain use cases necessitate the deduction of contributions from specific clients, a process known as “unlearning.” Existing machine unlearning methods are designed for centralized settings and do not cater to the collaborative nature of recommendation systems, thereby overlooking their unique characteristics. This article proposes CFRU, a novel federated recommendation unlearning model that enables efficient and certified removal of target clients from the global model. Instead of retraining the model, our approach rolls back and eliminates the historical updates associated with the target client. To efficiently store the learning process’s historical updates, we propose sampling strategies that reduce the number of historical updates, retaining only the most significant ones. Furthermore, we analyze the potential bias introduced by the removal of target clients’ updates at each training round and establish an estimation using the Lipschitz condition. Leveraging this estimation, we propose an efficient iterative scheme to accumulate the bias across all rounds, compensating for the removed updates from the global model and recovering its utility without requiring post-training steps. Extensive experiments conducted on two real-world datasets, incorporating two poison attack scenarios, have shown that our unlearning technique can achieve a model quality that is 99.3% equivalent to retraining the model from scratch while performing up to 1,000 times faster.

Multipath QUIC (MPQUIC) is an emerging multi-path transport protocol that lets a mobile client simultaneously use several wireless networks (e.g., Wi-Fi and cellular) in 5G and beyond. MPQUIC's performance heavily relies on its scheduler, which determines a path or several ones for sending packets in the upcoming time slot. Despite numerous efforts, the traditional design of MPQUIC schedulers can not handle wireless networks' dynamicity. Recently, a learning-based approach has shown the potential to bypass such limitations of the MPQUIC scheduler with various learning-based schedulers proposed in the literature. However, the existing works only consider the scheduling task in a single client context. When applying such a scheduler to multiple client scenarios (likely to occur in practice), they suffer from a so-called rush scheduling phenomenon. More specifically, the packet forwarding decisions made by a scheduler are only accountable to one client, resulting in conflicts of interest with other clients' schedulers. Consequently, it may harm the network performance. This paper addresses the issue and designs a learning-based MPQUIC scheduler considering the existence of multiple clients. To the best of our knowledge, this is the first work to do so. We propose MuLeS, a learning-based scheduler for MPQUIC in the multi-client scenario. MuLeS uses a central controller, which allows it to observe the state of all flows in the network. Our evaluation results show that MuLeS outperforms contemporary schedulers in terms of various metrics, including download time and loss rate. Notably, MuLeS reduces the average download time by 7%-16% compared to the other schedulers.

The increasing availability of multimodal data from electronic health records (EHR) has paved the way for deep learning methods to improve diagnosis accuracy.However, deep learning models are data-driven, requiring large-scale datasets to achieve high generalizability.Inspired by how human experts leverage reasoning for medical diagnosis, we propose CARER, a novel health risk prediction framework that enhances deep learning models with clinical rationales derived from medically proficient Large Language Models (LLMs).In addition, we provide a cross-view alignment loss which aligns the "local" view from the patient's health status with the "global" view from the external LLM's clinical reasoning to boost the mutual feature learning.Through extensive experiments on two predictive tasks using two popular EHR datasets, our CARER's significantly exceeds the performance of state-of-the-art models by up to 11.2 %, especially in improving data efficiency and generalizability 1 .

Reinforcement learning (RL) has been successfully applied in many fields for building autonomous systems, such as robotics and telecommunications. With its self-learning ability, RL provides a framework for learning from historical experience and adapting to dynamic environments. In response to the surge in network traffic and the evolving nature of traffic behavior, RL has emerged as a crucial technique for developing intelligent and adaptive traffic engineering (TE) solutions. However, most prior studies have focused on using a centralized unit (i.e., a single agent) to construct RL-based TE systems. While the centralized approach leverages global network information for solid performance, it encounters challenges related to scalability, dynamic network topology, and high monitoring overhead for collecting network information. This paper addresses these issues by introducing a jointly trained multi-agent reinforcement learning-based traffic engineering (MATE-JT) system, which operates as a distributed TE solution. Our approach utilizes multiple agents within a network node so that each agent can make independent routing decisions for a subset of flows. We take the approach of sharing parameters among agents and introduce a joint-training technique that facilitates simultaneous learning from multiple agents’ experiences. As a result, our proposed method enhances system performance while reducing training time. We evaluate the proposed approach using various network traffic datasets and demonstrate that MATE-JT improves the performance of TE (about 6.5%) and achieves faster convergence (about 35%) in large-scale networks when compared to state-of-the-art methods.

In this work, we investigate the time series representation learning problem using self-supervised techniques. Contrastive learning is well-known in this area as it is a powerful method for extracting information from the series and generating task-appropriate representations. Despite its proficiency in capturing time series characteristics, these techniques often overlook a critical factor - the inherent noise in this type of data, a consideration usually emphasized in general time series analysis. Moreover, there is a notable absence of attention to developing efficient yet lightweight encoder architectures, with an undue focus on delivering contrastive losses. Our work address these gaps by proposing an innovative training strategy that promotes consistent representation learning, accounting for the presence of noise-prone signals in natural time series. Furthermore, we propose an encoder architecture that incorporates dilated convolution within the Inception block, resulting in a scalable and robust network with a wide receptive field. Experimental findings underscore the effectiveness of our method, consistently outperforming state-of-the-art approaches across various tasks, including forecasting, classification, and abnormality detection. Notably, our method attains the top rank in over two-thirds of the classification UCR datasets, utilizing only 40% of the parameters compared to the second-best approach.

Offline optimization is an important task in numerous material engineering domains where online experimentation to collect data is too expensive and needs to be replaced by an in silico maximization of a surrogate of the black-box function. Although such a surrogate can be learned from offline data, its prediction might not be reliable outside the offline data regime, which happens when the surrogate has narrow prediction margin and is (therefore) sensitive to small perturbations of its parameterization. This raises the following questions: (1) how to regulate the sensitivity of a surrogate model; and (2) whether conditioning an offline optimizer with such less sensitive surrogate will lead to better optimization performance. To address these questions, we develop an optimizable sensitivity measurement for the surrogate model, which then inspires a sensitivity-informed regularizer that is applicable to a wide range of offline optimizers. This development is both orthogonal and synergistic to prior research on offline optimization, which is demonstrated in our extensive experiment benchmark.

Modern automobiles are equipped with advanced computing capabilities, allowing them to become powerful computing units capable of processing a large amount of data and training machine learning models. However, machine learning algorithms typically require a large centralized dataset, raising concerns about users’ privacy. Federated Learning (FL) is a distributed machine learning paradigm that tackles this problem, allowing intelligent vehicles to collaboratively train machine learning models locally without having to compromise their private data. Multiple works have concentrated on applying Federated Learning on Mobile Edge Computing (MEC) networks with a 3-tier architecture consisting of mobile clients, edge servers, and cloud servers, where the edge server aggregates its local set of clients, and the cloud server aggregates edge servers to learn a global model. This approach helps reduce the expensive communication costs to the far-away cloud server. However, this 3-tier paradigm faces several challenges, a notable one being clients’ constant mobility, leading to regional edges having a fluctuating set of participating clients at each round, which we refer to as distribution drift. This phenomenon introduces instability to the local training process, leading to suboptimal accuracy and convergence. As a solution, we propose a local training process based on the knowledge distillation mechanism. Specifically, we employ the global model and an ensemble of historical regional models from the edge servers as sources of knowledge to guide the local training process, preventing the local models from drifting away from the global knowledge and preserving information from clients that left the region. Experimental results showed that the proposed method helps achieve better performance compared to other baselines.

Federated Learning has emerged as a decentralized method for training machine learning models using distributed data sources. It ensures privacy by allowing clients to collaboratively learn a shared global model while keeping their data stored locally. However, a significant challenge arises when dealing with missing modalities in clients’ datasets, where certain features or modalities are unavailable or incomplete, leading to heterogeneous data distribution. Previous studies have addressed this issue, but they fall short in addressing generalizability across diverse, unobserved individuals. This study introduces MIFL, a novel framework for handling modality-missing clients in Multimodal Federated Learning. MIFL utilizes a unique approach, optimizing a client’s local model with existing modalities while incorporating absent modalities from clients. Aggregation of these models is performed through a graph-based attentive aggregation method, maintaining generalized characteristics by updating a global model averaged across clients. Our experimental results demonstrate the effectiveness of MIFL across various client configurations with statistical heterogeneity, showcasing its potential for addressing the challenge of missing modalities in Federated Learning.

Federated Learning has emerged as a transformative paradigm in the realm of collaborative machine learning, enabling the training of global models across decentralized devices without the need for centralizing data. While Federated Learning has shown remarkable promise, a critical limitation lies in its ability to personalize models to individual clients. Current approaches predominantly emphasize improving the accuracy of trained clients, inadvertently sidelining the significance of accommodating unseen clients. Furthermore, most of the existing personalized federated learning approaches require new clients to provide labeled data and undergo extensive retraining, posing a substantial barrier and hindering the broader adoption and engagement of potential users within these systems.In this paper, we introduce a novel and comprehensive solution to address these challenges: a generalized method for Personalized Federated Learning. Our approach transcends the limitations of conventional Federated Learning techniques by not only optimizing the accuracy of trained clients but also ensuring exceptional performance among unseen clients, even in diverse settings. Throughout extensive experiments, our method demonstrates significant improvements concerning the performance of seen and unseen clients, respectively, while eliminating the need for labeled data and model re-training among unseen clients.

Federated Learning (FL) is a method for training machine learning models using distributed data sources. It ensures privacy by allowing clients to collaboratively learn a shared global model while storing their data locally. However, a significant challenge arises when dealing with missing modalities in clients’ datasets, where certain features or modalities are unavailable or incomplete, leading to heterogeneous data distribution. While previous studies have addressed the issue of complete-modality missing1, they fail to tackle partial-modality missing2on account of severe heterogeneity among clients at an instance level, where the pattern of missing data can vary significantly from one sample to another. To tackle this challenge, this study proposes a novel framework named FedMAC, designed to address multimodality missing under conditions of partial-modality missing in FL. Additionally, to avoid trivial aggregation of multi-modal features, we introduce contrastive-based regularization to impose additional constraints on the latent representation space. The experimental results demonstrate the effectiveness of FedMAC across various client configurations with statistical heterogeneity, outperforming baseline methods by up to 26% in severe missing scenarios, highlighting its potential as a solution for the challenge of partially missing modalities in federated systems.1Complete missing is when one or more modalities are absent in server and clients’ data.2Partial missing is when only parts of one or more modalities are absent in server and clients’ data

Offline optimization has recently emerged as an increasingly popular approach to mitigate the prohibitively expensive cost of online experimentation. The key idea is to learn a surrogate of the black-box function that underlines the target experiment using a static (offline) dataset of its previous input-output queries. Such an approach is, however, fraught with an out-of-distribution issue where the learned surrogate becomes inaccurate outside the offline data regimes. To mitigate this, existing offline optimizers have proposed numerous conditioning techniques to prevent the learned surrogate from being too erratic. Nonetheless, such conditioning strategies are often specific to particular surrogate or search models, which might not generalize to a different model choice. This motivates us to develop a model-agnostic approach instead, which incorporates a notion of model sharpness into the training loss of the surrogate as a regularizer. Our approach is supported by a new theoretical analysis demonstrating that reducing surrogate sharpness on the offline dataset provably reduces its generalized sharpness on unseen data. Our analysis extends existing theories from bounding generalized prediction loss (on unseen data) with loss sharpness to bounding the worst-case generalized surrogate sharpness with its empirical estimate on training data, providing a new perspective on sharpness regularization. Our extensive experimentation on a diverse range of optimization tasks also shows that reducing surrogate sharpness often leads to significant improvement, marking (up to) a noticeable 9.6% performance boost. Our code is publicly available at https://github.com/cuong-dm/IGNITE.

Over recent decades, air pollution, particularly fine particulate matter (PM2.5), has garnered significant attention due to its detrimental effects on human health. As a result, accurately forecasting air quality has become a critical objective to inform and guide policy-making. However, in developing countries, due to their financial constraints, the prediction ability mostly depends on sparsely distributed monitoring devices, which are insufficient. Various prediction approaches utilizing deep learning techniques, such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models, have been developed to address this gap. Despite their promise, these models encounter challenges, including slow preprocessing times and limited adaptability to changing air quality conditions. This paper proposes a simple Encoder-Decoder model for predicting hourly and daily PM2.5 concentrations. In addition, this paper implements an automated system that collects and processes real-time data, continuously updating the model to respond to emerging trends and variations in pollution levels. The model’s performance is evaluated in Vietnam’s two largest cities, Hanoi and Ho Chi Minh City, demonstrating prediction accuracy in various scenarios.

With the increasing demand for video streaming services, efficient video streaming with MPQUIC in mobile wireless networks is gaining interest. Although MPQUIC can leverage multiple network connections (e.g., Wi-Fi, 5G) for simultaneous transfer, improving bandwidth and resilience, its performance is heavily dependent on the MPQUIC scheduler. In mobile scenarios, network fluctuations challenge the scheduler to adapt to dynamic conditions while maintaining good video streaming quality. Addressing this issue, this paper proposes a novel MPQUIC scheduler named LoGra (i.e., LSTM-DDPG integrated MPQUIC Scheduler) with two advanced features. First, LoGra utilizes Long Short-Term Memory (LSTM) to model the temporal correlation of network conditions over time and handle the challenges posed by mobility patterns. Second, it leverages the Deep Deterministic Policy Gradient (DDPG), with its self-learning capabilities and the strength of deep neural networks, to analyze the informative temporal feature vector to influence scheduling decisions. We have implemented the proposed scheduler and compared it to existing ones. The results show that the LoGra scheduler effectively manages multipath communication in heterogeneous wireless networks under mobility scenarios. Moreover, compared to existing schedulers, LoGra achieves significant improvements in data transmission, both in general metrics (loss, goodput) and streaming-related metrics (bitrate, freezing time).

Due to global air pollution , there is a growing demand for accurate and large-scale air quality monitoring systems. Consequently, low-cost air monitoring devices have emerged as a potential alternative to expensive conventional ones. However, the low-cost devices’ major drawback is their insufficient level of accuracy. This work investigates the problem of calibrating the sensory data, especially PM2.5 concentration, collected by low-cost sensor-based air quality monitoring devices. Recently, deep learning has emerged as a potential solution for data calibration instead of using traditional methods, whose accuracy is relatively low. Nevertheless, it generally incurs significant costs. Moreover, it is necessary to employ a dedicated calibration model for each device to increase precision, resulting in additional expenditures. To address the issue, this study provides a novel approach named GAMMA, which entails the development of a deep learning-based model capable of accurately calibrating data for multiple devices simultaneously. The proposed method leverages the multitask learning paradigm to solve the challenge of concurrently processing several devices’ data. This involves capturing common features across all devices’ data while also distinguishing the device-specific characteristics. Furthermore, GAMMA also employs the adversarial training approach to augment the accuracy of predictions. This method has been implemented and integrated into an air quality monitoring system in Hanoi, Vietnam. Comprehensive experiments are conducted on real-world data to demonstrate the superiority of GAMMA against the comparison benchmarks in terms of various metrics. Notably, GAMMA reduces MAE from 60.19% to 74.09% compared to the best comparison baseline. The source code is available at https://github.com/anhduy0911/FimiCalibIdea/tree/multi_attention .

Recent advancements in recommender systems have focused on integrating knowledge graphs (KGs) to leverage their auxiliary information. The core idea of KG-enhanced recommenders is to incorporate rich semantic information for more accurate recommendations. However, two main challenges persist: i) Neglecting complex higher-order interactions in the KG-based user-item network, potentially leading to sub-optimal recommendations, and ii) Dealing with the heterogeneous modalities of input sources, such as user-item bipartite graphs and KGs, which may introduce noise and inaccuracies. To address these issues, we present a novel Knowledge-enhanced Heterogeneous Hypergraph Recommender System (KHGRec). KHGRec captures group-wise characteristics of both the interaction network and the KG, modeling complex connections in the KG. Using a collaborative knowledge heterogeneous hypergraph (CKHG), it employs two hypergraph encoders to model group-wise interdependencies and ensure explainability. Additionally, it fuses signals from the input graphs with cross-view self-supervised learning and attention mechanisms. Extensive experiments on four real-world datasets show our model's superiority over various state-of-the-art baselines, with an average 5.18% relative improvement. Additional tests on noise resilience, missing data, and cold-start problems demonstrate the robustness of our KHGRec framework. Our model and evaluation datasets are publicly available at https://github.com/viethungvu1998/KHGRec.

Traffic engineering (TE) is one of the most critical issues in networking, as it enables efficient and reliable network operations. With the advent of Machine Learning (ML) techniques, many ML-based TE methods have emerged in recent years, especially those employing Deep Neural Networks for future traffic prediction to enhance the performance of traditional approaches. However, current methods suffer from two major issues. Firstly, most prior works only solve the TE problem based on short-term traffic prediction, neglecting the network traffic dynamics over an extended time period. This oversight results in high network disturbance when numerous traffic flows need to be rerouted to adapt to traffic changes. Secondly, although traffic prediction models rely on historical traffic data to perform future prediction, ML-based TE studies often ignore the high overhead for network traffic monitoring. To address these issues, we propose a traffic prediction-based routing algorithm in which the routing rules can be applied to multiple time-steps without requiring changes, ultimately leading to reduced network disturbance. We employ the segment routing (SR) technique as the routing algorithm and formulate the multi-time-step segment routing method that incorporates future traffic prediction. To address the high monitoring overhead, we present an approach that combines partial traffic prediction and compressive sensing techniques to estimate unmeasured data. Through extensive experiments on real backbone network traffic datasets, we demonstrate that our proposal can achieve more than 80% of the optimal performance in reducing maximum link utilization while significantly reducing the number of routing changes and traffic monitoring cost.

In Federated Learning (FL), the size of local models matters. On the one hand, it is logical to use large-capacity neural networks in pursuit of high performance. On the other hand, deep convolutional neural networks (CNNs) are exceedingly parameter-hungry, which makes memory a significant bottleneck when training large-scale CNNs on hardware-constrained devices such as smartphones or wearables sensors. Current state-of-the-art (SOTA) FL approaches either only test their convergence properties on tiny CNNs with inferior accuracy or assume clients have the adequate processing power to train large models, which remains a formidable obstacle in actual practice. To overcome these issues, we introduce FedDCT, a novel distributed learning paradigm that enables the usage of large, high-performance CNNs on resource-limited edge devices. As opposed to traditional FL approaches, which require each client to train the full-size neural network independently during each training round, the proposed FedDCT allows a cluster of several clients to collaboratively train a large deep learning model by dividing it into an ensemble of several small sub-models and train them on multiple devices in parallel while maintaining privacy. In this collaborative training process, clients from the same cluster can also learn from each other, further improving their ensemble performance. In the aggregation stage, the server takes a weighted average of all the ensemble models trained by all the clusters. FedDCT reduces the memory requirements and allows low-end devices to participate in FL. We empirically conduct extensive experiments on standardized datasets, including CIFAR-10, CIFAR-100, and two real-world medical datasets HAM10000 and VAIPE. Experimental results show that FedDCT outperforms a set of current SOTA FL methods with interesting convergence behaviors. Furthermore, compared to other existing approaches, FedDCT achieves higher accuracy and substantially reduces the number of communication rounds (with 4-8 times fewer memory requirements) to achieve the desired accuracy on the testing dataset without incurring any extra training cost on the server side.

Federated learning enables edge devices to train a global model collaboratively without exposing their data. Despite achieving outstanding advantages in computing efficiency and privacy protection, federated learning faces a significant challenge when dealing with non-IID data, i.e., data generated by clients that are typically not independent and identically distributed. In this paper, we tackle a new type of Non-IID data, called cluster-skewed non-IID, discovered in actual data sets. The cluster-skewed non-IID is a phenomenon in which clients can be grouped into clusters with similar data distributions. By performing an in-depth analysis of the behavior of a classification model's penultimate layer, we introduce a metric that quantifies the similarity between two clients' data distributions without violating their privacy. We then propose an aggregation scheme that guarantees equality between clusters. In addition, we offer a novel local training regularization based on the knowledge-distillation technique that reduces the overfitting problem at clients and dramatically boosts the training scheme's performance. We theoretically prove the superiority of the proposed aggregation over the benchmark FedAvg. Extensive experimental results on both standard public datasets and our in-house real-world dataset demonstrate that the proposed approach improves accuracy by up to 16% compared to the FedAvg algorithm.

In the era of 5G and beyond, mobile devices usually can access several heterogeneous wireless networks (e.g., Wi-Fi and 5G). To simultaneously and efficiently utilize the accessible network resources, muli path transport protocols, such as MPTCP and MPQUIC, have shown much potential. In these protocols, scheduling is one of the critical processes to ensure the performance of the multipath transmission. Although there have been many proposed multipath schedulers in the literature, they have not performed well in heterogeneous networks, especially when the network conditions vary (i.e., dynamicity). In this paper, we propose a novel Q-learning-based Multipath scheduler for data transmission optimization (Q-SAT), aiming to bypass the existing limitation. By leveraging the self-learning ability of reinforcement learning, Q-SAT can instantly observe environmental changes and deploy appropriate path selection to optimize data transmission time. As a result, Q-SAT efficiently schedules multipath communication in heterogeneous wireless networks with different dynamicity levels. We have implemented Q-SAT with MPQUIC and extensively evaluated Q-SAT in an emulated environment and a real network. The evaluation results show that Q-SAT improves the data transmission time by at least 10% in the emulation and 26% in the actual deployment compared to the state-of-the-art schedulers.

Federated learning (FL) enables multiple clients to train a model without compromising sensitive data. The decentralized nature of FL makes it susceptible to adversarial attacks, especially backdoor insertion during training. Recently, the edge-case backdoor attack employing the tail of the data distribution has been proposed as a powerful one, raising questions about the shortfall in current defenses' robustness guarantees. Specifically, most existing defenses cannot eliminate edge-case backdoor attacks or suffer from a trade-off between backdoor-defending effectiveness and overall performance on the primary task. To tackle this challenge, we propose FedGrad, a novel backdoor-resistant defense for FL that is resistant to cutting-edge backdoor attacks, including the edge-case attack, and performs effectively under heterogeneous client data and a large number of compromised clients. FedGrad is designed as a two-layer filtering mechanism that thoroughly analyzes the ultimate layer's gradient to identify suspicious local updates and remove them from the aggregation process. We evaluate FedGrad under different attack scenarios and show that it significantly outperforms state-of-the-art defense mechanisms. Notably, FedGrad can almost 100% correctly detect the malicious participants, thus providing a significant reduction in the backdoor effect (e.g., backdoor accuracy is less than 8%) while not reducing main accuracy on the primary task.

Advances in deep neural network (DNN) architectures have enabled new prediction techniques for stock market data. Unlike other multivariate time-series data, stock markets show two unique characteristics: (i) multi-order dynamics, as stock prices are affected by strong non-pairwise correlations (e.g., within the same industry); and (ii) internal dynamics, as each individual stock shows some particular behaviour. Recent DNN-based methods capture multi-order dynamics using hypergraphs, but rely on the Fourier basis in the convolution, which is both inefficient and ineffective. In addition, they largely ignore internal dynamics by adopting the same model for each stock, which implies a severe information loss.

With recent advancements in graph neural networks (GNN), GNN-based recommender systems (gRS) have achieved remarkable success in the past few years. Despite this success, existing research reveals that gRSs are still vulnerable to poison attacks , in which the attackers inject fake data to manipulate recommendation results as they desire. This might be due to the fact that existing poison attacks (and countermeasures) are either model-agnostic or specifically designed for traditional recommender algorithms (e.g., neighborhood-based, matrix-factorization-based, or deep-learning-based RSs) that are not gRS. As gRSs are widely adopted in the industry, the problem of how to design poison attacks for gRSs has become a need for robust user experience. Herein, we focus on the use of poison attacks to manipulate item promotion in gRSs. Compared to standard GNNs, attacking gRSs is more challenging due to the heterogeneity of network structure and the entanglement between users and items. To overcome such challenges, we propose GSPAttack —a generative surrogate-based poison attack framework for gRSs. GSPAttack tailors a learning process to surrogate a recommendation model as well as generate fake users and user-item interactions while preserving the data correlation between users and items for recommendation accuracy. Although maintaining high accuracy for other items rather than the target item seems counterintuitive, it is equally crucial to the success of a poison attack. Extensive evaluations on four real-world datasets revealed that GSPAttack outperforms all baselines with competent recommendation performance and is resistant to various countermeasures.

This paper introduces a host-based method for investigating IPv6 adoption and performance in academia that targets universities in Japan and Vietnam. Unlike other works, the investigation has been utilized on a host native IPv6 host with a standard tool (i.e., curl). We first probe the IPv6 capabilities of the universities’ websites. Second, within the IPv6-supported websites, we compare the performances of IPv4 and IPv6 when letting the two IP versions access the websites concurrently. We find that, despite the popularity of IPv6 in the two countries, a significant number of academic websites are not yet IPv6 capable. The native IPv6 client can only access fifty of the more than one thousand websites in Japan. Furthermore, there are no IPv6-supported websites at Vietnamese universities. Among the accessible IPv6 websites in Japan, we observe that the web access performances of IPv4 and IPv6 are similar.

The uneven distribution of local data across different edge devices (clients) results in slow model training and accuracy reduction in federated learning. Naive federated learning (FL) strategy and most alternative solutions attempted to achieve more fairness by weighted aggregating deep learning models across clients. This work introduces a novel non-IID type encountered in real-world datasets, namely cluster-skew, in which groups of clients have local data with similar distributions, causing the global model to converge to an over-fitted solution. To deal with non-IID data, particularly the cluster-skewed data, we propose FedDRL, a novel FL model that employs deep reinforcement learning to adaptively determine each client’s impact factor (which will be used as the weights in the aggregation process). Extensive experiments on a suite of federated datasets confirm that the proposed FedDRL improves favorably against FedAvg and FedProx methods, e.g., up to 4.05% and 2.17% on average for the CIFAR-100 dataset, respectively.

Target coverage and connectivity are two of the most crucial issues in handling wireless sensor networks. However, maintaining these two factors is challenging due to the energy constraint of sensors. To this end, wireless charging has emerged as a promising solution to prolong the sensor's lifetime. In a wireless charging sensor network, a mobile charger moves around the network, stops at several charging locations and charges the sensor via electromagnetic waves. In this study, we investigate the problem of optimizing the charging location and charging time of the mobile charger to ensure the target coverage and connectivity of the network. Our main idea is to leverage the Deep Reinforcement Learning approach. Specifically, the mobile charger will act as an agent, which receives a state including the energy information of the sensors. The mobile charger then decides the following charging location and charging time using the state information and the knowledge learned in the past. Experimental results have shown that our algorithm can extend the network lifetime (i.e., the time until the network coverage and connectivity are not guaranteed) up to 245.9 times compared to the existing algorithms.

In this paper, we address vehicle-based mobile crowdsensing for air quality monitoring applications. We tackle a novel issue that asks to determine monitoring frequencies for maximizing spatial-temporal coverage while reducing the monitoring costs and balancing load across the vehicles. We begin by theoretically formulating the problem and proposing an objective function that considers the three goals. We then leverage the evolutionary approach to develop an algorithm for determining the optimal monitoring frequency. We conduct comprehensive experiments to evaluate the performance of the proposed approach and compare it to the other methods. The results indicate that our approach can enhance the objective function by a factor of 1.33 to 4 compared to the others.

Multi-access edge computing (MEC) is seen as an effective technique for decreasing service latency in a V2X network by offloading computational activities. With MEC, a three-tier offloading architecture can be developed, where a vehicle can offload computational tasks to a cloud by communicating with a base station (gNB) and Road Side Units (RSUs). In this paper, we focus on three-tier V2X networks which rely on three offloading paths: Vehicle-to-Infrastructure (i.e., vehicle to RSU), Vehicle-to-Cloud (i.e., vehicle to gNB), and Infrastructure-to-Cloud (i.e., RSU to gNB). We propose an offloading strategy based on deep reinforcement learning with the goal of reducing the average latency of tasks. To be more specific, we leverage the Deep Q Network to estimate the goodness of action-state value to determine the offloading decision. We also propose a novel exploration scheme and a new model training strategy. The experimental findings indicate that our proposed offloading method outperforms the state-of-the-art, particularly in critical circumstances characterized by a high rate of vehicle arrival or packet generation.

This study focuses on MEC-enhanced, vehicle-based crowdsensing systems that rely on devices installed on automobiles. We investigate an opportunistic communication paradigm in which devices can transmit measured data directly to a crowdsensing server over a 4G communication channel or to nearby devices or so-called Road Side Units positioned along the road via Wi-Fi. We tackle a new problem that is how to reduce the cost of 4G while preserving the latency. We propose an offloading strategy that combines a reinforcement learning technique known as Q-learning with Fuzzy logic to accomplish the purpose. Q-learning assists devices in learning to decide the communication channel. Meanwhile, Fuzzy logic is used to optimize the reward function in Q-learning. The experiment results show that our offloading method significantly cuts down around 30-40% of the 4G communication cost while keeping the latency of 99% packets below the required threshold.

In a Wireless Rechargeable Sensor Network (WRSN), a mobile charger (MC) moves and supplies energy for sensor nodes to maintain the network operation. Hence, optimizing the charging schedule of MC is essential to maximize the network lifetime in WRSNs. The existing works only target the local optimization of network lifetime limited to MC’s subsequent charging round. The network lifetime has been normally reflected in a different metric that is not directly related to the final charging round period. To the best of our knowledge, this work is the first to address the global maximization of network lifetime in WRSNs, which optimizes not only the subsequent charging round but all charging rounds over the entire network lifetime. Another uniqueness is the joint consideration of both the charging path and charging time optimization problems. As a solution, we propose a genetic algorithm (GA)-based global optimization scheme that considers all the possible charging rounds. The GA has a novel mutation operation that mutates gene sizes for representing charging schedules with a varying number of charging rounds. The experiment results show that our algorithm can extend the network lifetime by 35.1 times on average and 38.6 times in the best case compared to existing ones.

Air pollution is a critical issue in cities in developing countries like Hanoi, Vietnam. An efficient and comprehensive air quality monitoring system may reduce the harmfulness and improve the cities' sustainability. This paper presents a novel approach to realize such a system in which the air monitoring sensors are mobile. More specifically, we introduce a three-tier architecture for the air quality system, including sensing, communication, and application layers. Initially, we discuss each layer concept to bypass the limitation of the traditional stationary monitoring system. We then describe our design and implementation of air quality monitoring devices installed on vehicles, such as buses. The device is carefully designed to satisfy the conditions of impedance matching and power integrity. Besides, it fully functions in measuring parameters from the ambient environment. The device is aware of its location (using GPS) and uses Wi-Fi and 4G (LTE) to transmit sensing data on the Internet. We have conducted various experiments, including a trial deployment of the devices on a vehicle running in Hanoi. The results show our device achieves sensing data transmission with high-reliability levels (i.e., 97%, 100% on Wi-Fi, 4G (LTE), respectively). Moreover, the trial deployment confirms the feasible operation of our device in actual condition.

In a three-tier Vehicle to X (V2X) network, a vehicle can offload the computational tasks to the edge computing component at a roadside unit (RSU) or a base station with cloud computing (gNB). Moreover, an RSU can also offload to gNB, forming three offloading paths: vehicle-to-RSU, vehicle-to-gNB, and RSU-to-gNB. This paper aims to minimize the offloaded tasks' average latency while dealing with the network dynamic. Note that the existing works assume the fixed network parameters, hence have failed to address the dynamic. As a solution, we use the multi-agent multi-armed bandits (MBA) learning for offloading that can adapt to the network dynamic and optimize the latency. More importantly, we propose a new MBA offloading scheme with an exploration mechanism based on the Sigmoid function. We conduct an extensive evaluation to evaluate and show the superiority of our proposal. First, the proposed Sigmoid exploration mechanism reduces the tasks' average latency by 35% compared to a basic MBA using negative rewarding. Second, the simulation results show our proposed offloading algorithm shortens the task latency by 18.5% on average and 56.9% in the best case, compared to the state-of-the-art.

Missing values appear in most multivariate time series, especially in the monitored network traffic data due to high measurement cost and unavoidable loss. In the networking fields, missing data prevents advanced analysis and downgrades downstream applications such as traffic engineering and anomaly detection. Despite the great potential, existing imputation approaches based on tensor decomposition and deep learning techniques have shown limitations in addressing missing values of traffic data due to its dynamic behavior. In this paper, we propose Graph Convolutional Recurrent Neural Network for Imputing Network Traffic (GCRINT), a combination between Recurrent Neural Network (RNN) and Graph Convolutional Neural Network, for filling the missing values of network traffic data. We use a bidirectional Long Short-Term Memory network and Graph Neural Network to efficiently learn the spatial-temporal correlations in partially observed data. We conducted extensive experiments to evaluate our model by using two different datasets and various missing scenarios. The experiment results show that GCRINT achieves significantly low imputation errors and reduces the error by 35% compared to the state-of-the-art methods. GCRINT also helps to obtain a stable performance in the traffic engineering problem.

In the age of the ever-growing number of tasks being generated from IoT devices, one of the most crucial problems with enhancing the Quality of Service in multi-access computing is the system's limited resources. To this end, Vehicular Fog Computing (VFC) has emerged as a potential solution that utilizes the idle resources of vehicles to reduce the load imposed on the edge servers. In this paper, we leverage the advantages of both deep reinforcement learning and Fuzzy logic to propose Fuzzy Deep Q-learning base Offloading scheme (FDQO), a real-time offloading scheme in delay constrained VFC. Our objective is to maximize the Quality of Experiences (QoE), which indicates how the task meets its delay constraint. The experiment results show that our proposed approach significantly outperforms the existing algorithms. Specifically, FDQO improves the average QoE by 37.72% compared to using only Deep Q-learning, 7.47% compared to using only Fuzzy logic, and 19% compared to the ∊ -greedy strategy for multi-armed bandits.

We focus on real-time air quality monitoring systems that rely on devices installed on automobiles in this research. We investigate an opportunistic communication model in which devices can send the measured data directly to the air quality server through a 4G communication channel or via Wi-Fi to adjacent devices or the so-called Road Side Units deployed along the road. We aim to reduce 4G costs while assuring data latency, where the data latency is defined as the amount of time it takes for data to reach the server. We propose an offloading scheme that leverages Q-learning to accomplish the purpose. The experiment results show that our offloading method significantly cuts down around 40-50% of the 4G communication cost while keeping the latency of 99.5% packets smaller than the required threshold.

With the recent explosive growth of mobile devices such as smartphones or tablets, guaranteeing consistent web appearance across all environments has become a significant problem. This happens simply because it is hard to keep track of the web appearance on different sizes and types of devices that render the web pages. Therefore, fixing the inconsistent appearance of web pages can be difficult, and the cost incurred can be huge, e.g., poor user experience and financial loss due to it. Recently, automated web repair techniques have been proposed to automatically resolve inconsistent web page appearance, focusing on improving usability. However, generated patches tend to disrupt the webpage's layout, rendering the repaired webpage aesthetically unpleasing, e.g., distorted images or misalignment of components. In this paper, we propose an automated repair approach for web pages based on meta-heuristic algorithms that can assure both usability and aesthetics. The key novelty that empowers our approach is a novel fitness function that allows us to optimistically evolve buggy web pages to find the best solution that optimizes both usability and aesthetics at the same time. Empirical evaluations show that our approach is able to successfully resolve mobile-friendly problems in 94% of the evaluation subjects, significantly outperforming state-of-the-art baseline techniques in terms of both usability and aesthetics.

The fifth generation of mobile wireless networks (5G) will provide an infrastructure with abundant and reliable connectivity for innovative and complicated applications. In 5G, 5G mobile devices, which will have improved computing resources for such applications, play an essential role. However, the network stack of 5G devices may continue to be borrowed from 4G legacy operating systems, thereby degrading user experience. In this paper, we posit that 5G users should have flexibility in utilizing networks and gain more awareness of network selection. To this end, we exploit network softwarization technologies to empower 5G devices. We then devise 5GSoft, a novel softwarized networking stack on each 5G mobile device. 5GSoft includes wireless virtualization to relax the dependence on hardware, thereby enabling sharing and multiple access. The 5GSoft device can concurrently exploit surrounding wireless networks using software-defined networking. Finally, the 5GSoft device is aware of network selection for each application process by applying network namespace. Qualitative evaluation of 5GSoft, in comparison to other approaches, highlights its effectiveness in terms of awareness and flexibility provision. Moreover, real experiments show that the virtualization in 5GSoft has negligible overhead.

Abstract A quest for geographic routing schemes of wireless sensor networks when sensor nodes are deployed in areas with obstacles has resulted in numerous ingenious proposals and techniques. However, there is a lack of solutions for complicated cases wherein the source or the sink nodes are located close to a specific hole, especially in cavern-like regions of large complex-shaped holes. In this paper, we propose a geographic routing scheme to deal with the existence of complicated-shape holes in an effective manner. Our proposed routing scheme achieves routes around holes with the (1+ $$\epsilon$$ ϵ )-stretch. Experimental results show that our routing scheme yields the highest load balancing and the most extended network lifetime compared to other well-known routing algorithms as well.

In recent years, Wireless Rechargeable Sensor Networks, which exploit wireless energy transfer technologies to address the energy constraint problem in traditional Wireless Sensor Networks, has emerged as a promising solution. There are two important factors that affect the performance of a charging process: charging path and charging time. In the literature, many studies have been done to propose efficient charging algorithms. However, most of the existing works focus only on optimizing the charging path. In this paper, we are the first one to jointly take into account both the charging path and charging time. Specifically, we aim at determining the optimal charging path and the charging time at each charging location to minimize the number of dead nodes. We first mathematically formulate the problem under mixed integer and linear programming. Then, we propose a periodic charging scheme, which is based on the Greedy and Genetic algorithm approaches. The experiment results show that our proposed the algorithm reduces significantly the number of dead nodes compared to a relevant benchmark.

Leveraging mobile cloud computing (MCC) and mobile edge computing (MEC) for offloading computational tasks is a promising approach to enabling delay-sensitive applications executing vehicles. Despite MCC and MEC's ability and complementary characteristics, most of the existing works on offloading focus on only either MCC or MEC. In this paper, we study their cooperation in a three-tier offloading model of a V2X network where a vehicle can offload computational tasks to cloud computing and MEC. Specifically, we investigate the optimal offloading probabilities of three offloading paths, including Vehicle-to- Infrastructure, Vehicle-to-Cloud, and Infrastructure-to-Cloud. Our contribution is twofold. First, we derive a mathematical model of task execution latency and a formulation to find an optimal solution for the minimum latency problem. Second, we propose an approximation algorithm based on the genetic algorithm toward the optimum. The experiment results show that by exploiting both MCC and MEC's complementary advantages, our proposed algorithm in the three-tier model can shorten the delay significantly compared to existing two-tier models. Depending on the traffic load and the number of Road Side Units, our proposal can reduce the delay by 93.75% on the average, and 99.9% in the best case.

So far, air quality monitoring is usually handled by monitoring stations located at fixed locations. However, due to the cost of installation, deployment, and operation, the number of monitoring stations deployed is often tiny; thus, the monitored area is limited. To deal with this problem, in this paper, we consider a mobile air quality monitoring system that relies on sensors mounted on buses to broaden the monitoring area. Specifically, we investigate the optimal buses to place the sensors as well as the optimal monitoring timings to maximize the number of critical regions that are monitored. We mathematically formulate the targeted problem and prove its NP-hardness. Then, we exploit the greedy and dynamic programming approaches to propose a polynomial-time 1/2-approximation algorithm. We use the data of real bus routes in Hanoi, Vietnam, for the experimentation and show that the proposed algorithm guarantees an average performance ratio of 72.68%.

In the last decades, scene text recognition has gained worldwide attention from both the academic community and actual users due to its importance in a wide range of applications. Despite achievements in optical character recognition, scene text recognition remains challenging due to inherent problems such as distortions or irregular layout. Most of the existing approaches mainly leverage recurrence or convolution-based neural networks. However, while recurrent neural networks (RNNs) usually suffer from slow training speed due to sequential computation and encounter problems as vanishing gradient or bottleneck, CNN endures a trade-off between complexity and performance. In this paper, we introduce SAFL, a self-attention-based neural network model with the focal loss for scene text recognition, to overcome the limitation of the existing approaches. The use of focal loss instead of negative log-likelihood helps the model focus more on low-frequency samples training. Moreover, to deal with the distortions and irregular texts, we exploit Spatial TransformerNetwork (STN) to rectify text before passing to the recognition network. We perform experiments to compare the performance of the proposed model with seven benchmarks. The numerical results show that our model achieves the best performance.

Mobile edge computing (MEC) is a model that allows integration of computing power into telecommunications networks, to improve communication and data processing efficiency. In general, providing power to ensure the computing power of edge servers in the MEC network is very important. In many cases, ensuring continuous power supply to the system is not possible because servers are deployed in hard-to-reach areas such as outlying areas, forests, islands, etc. This is when renewable energy prevails as a viable source of power for ensuring stable operation. This paper addresses resource provisioning in the MEC network using renewable energy. We formulate the problem as a Markov Decision Problem and introduce a new approach to optimize this problem in terms of energy and time costs by using a reinforcement learning technique. Our simulation validates the efficacy of our algorithm, which results in a cost three times better than the other methods.

In this paper, we focus on broadening the monitoring area of a mobile air quality monitoring system, in which the sensors mounted on buses. In particular, we investigate the optimal buses to place the sensors and the optimal monitoring timings to maximize the number of monitored critical regions. We mathematically formulate the targeted problem. Then, we leverage the greedy approach to propose a polynomial-time$\frac{e-1}{2e-1}$approximation algorithm. We use the data of real bus routes in Hanoi, Vietnam, for the experimentation and show that the proposed algorithm guarantees an average performance ratio of 63.87%.

This paper introduces a novel charging strategy for wireless rechargeable sensor networks (WRSNs), in which a mobile charger (MC) moves and wirelessly transfers the power to the sensor nodes. The first distinct point of this work is designing the MC's charging algorithm under the consideration of target coverage and connectivity. As a solution, we introduce a novel on-demand charging scheme for WRSNs that optimize the charging time at each MC's charging location. Moreover, we take advantage of the Q-learning technique (i.e., hence named our algorithms Q-charging) to maximize the number of monitored targets. Q-charging can prioritize the sensor nodes, which play a more critical role in the network. Hence, Q-charging can select a suitable charging location aiming to provide sufficient power for the prioritized sensors. We have evaluated our proposal in comparison to the previous works. The evaluation results show that Q-charging can prolong the time until the first target is not monitored by 5.2 times on the average, and 14.3 times in the best case, compared to existing algorithms.

Thanks to the advancements in wireless power transfer technologies, a new paradigm of wireless sensor network (WSNs) called wireless rechargeable sensor networks (WRSNs) has recently emerged. For a WRSN, designing an efficient charging schedule is a challenging issue due to the inherent constraints of WSNs. Although there have been many efforts to optimize the charging schedule, the existing works suffer from several critical problems. Firstly, they rarely tackle the dead node minimization problem, which is the ultimate objective of wireless charging. Secondly, most of the existing works assume impractical conditions, which include the unlimited battery capacity of the charger, and a fully charging scheme at the sensors. In this paper, aiming at minimizing the number of dead nodes, we propose a novel charging scheme based on the genetic algorithm. Our scheme works when the mobile charger has only limited capacity, and the sensors are charged partially at each charging round. The experiment results show that our proposed algorithm reduces the number of dead nodes significantly compared to other existing studies.

Sensor scheduling for maximizing the network lifetime and achieving the full area coverage is a paramount problem in wireless sensor networks. Although considerable effort has been devoted, this problem is still a challenge to the research community. The approximation algorithms proposed so far couldn't guarantee the performance ratio. In this paper, we first formulate the problem under linear programming model which can help to determine the exact optimal solution. Then, in order to reduce the time complexity, we propose a (1 +∊)-approximation algorithm based on divide-and-conquer technique. The main idea is to divide the network into sub-regions, then determine the suboptimal solution for every sub-region and combine them to obtain the total solution of the whole network. Moreover, with the aim of speeding up the suboptimal solution finding process, we propose an approximation algorithm using the column generation approach. The experiment results show the superiority of our proposed algorithms over the existing ones.

Geographic routing is one of the most popular routing protocols in wireless sensor networks (WSNs) due to its simplicity and efficiency. However, with the occurrence of holes, geographic routing incurs with the so-called local minimum problem that may lead to a long hole detour path as well as the traffic concentration around the hole boundary. In consequence, the network lifetime is shortened. In this paper, we aim at proposing a lightweight distributed geographic routing protocol, which can prolong the lifetime of WSNs under the hole occurrence. Our main idea is to exploiting Q-learning technique to estimate the distance from a node to the holes. The routing decision is then determined based on the residual energy of the nodes, their estimated distance to the holes, and their distance to the destination. The simulation experiments show that our protocol strongly outperforms state-of-the-art protocols in terms of the network lifetime, packet latency and energy consumption. Specifically, our proposed protocol extends the network lifetime by more than 12% compared to the existing protocols.

Because of its simplicity and scalability, geographic routing is a popular approach in wireless sensor networks, which can achieve a near-optimal routing path in the networks without holes (i.e., regions without working sensors). With the occurrence of holes, however, geographic routing faces the problems of load imbalance and routing path enlargement. In the literature, several proposals have attempted to fix these issues, but the majority of them considers only the cases when both the source and the destination stay fairly far from the holes. Recently, a few work has been proposed to tackle the problem of routing in the vicinity of routing holes. However, none of them addresses the two problems (i.e., load imbalance and routing path enlargement) concurrently, and none of them can solve the problem of load imbalance thoroughly. In this paper, we introduce a novel approach in dealing with routing in the vicinity of holes, that is the first to target and solve both the load imbalance and path enlargement problems. The theoretical analysis proves that the routing path stretch of our proposed protocol can be controlled to be as small as 1 + ε (for any predefined ε> 0) and the simulation experiments show that our protocol strongly outperforms the existing protocols in terms of load balancing.

Target coverage and connectivity are two main challenging and important issues in wireless sensor networks. The former is for providing sufficient monitoring quality where all points of interest in the network are covered by sensor nodes and the latter is for guaranteeing satisfactory communicating capability where all sensors can connect to at least one sink via relay nodes. In this paper, we focus on minimizing the number of nodes (i.e., sensor nodes and relay nodes) to provide target coverage and connectivity in wireless sensor networks with multiple sinks. We formulate the problem as two sub-problems. The first one (named as TC) is for placing sensor nodes to cover all targets and the second one (named as NC) is for placing relay nodes to connect sensor nodes to the sinks. We then propose a heuristic algorithm for the TC problem that exploits clustering technique. We also propose two heuristic algorithms for the NC problem that base on greedy approach and spanning tree. The experiment results show that our protocols can significantly reduce the number of required nodes in comparison with existing protocols.

Wireless sensor networks (WSNs) are used in many mission-critical applications, such as target tracking on a battlefield, emergency alarms, and disaster detection. In such applications, QoS provisioning in the timeliness domain is indispensable. Moreover, because of the diversity of sensory data, QoS provisioning should support not only one but multiple levels of end-to-end delay constraints. As a result of several characteristics such as the limitations on the energy supply, available storage and computational capacity of the sensor nodes, guaranteeing timely delivery in WSNs is a challenging problem. To overcome these limitations, several lightweight and stateless QoS-based geographic routing protocols have been proposed. The existing protocols work well in networks without routing holes (i.e., regions with no working sensors). However, with the occurrence of routing holes, they suffer from the so-called local minimum phenomenon and traffic congestion around the hole boundary. In this paper, we consider the presence of routing holes and propose a delay-guaranteed geographic routing protocol called DEHA that can support multiple end-to-end delay levels. The main idea is to achieve early awareness of the presence of a routing hole and then to utilize this awareness in determining a routing path that can avoid the hole. Simulation results show that our protocol outperforms the existing protocols in terms of several performance metrics, including packet delivery ratio, energy efficiency, and load balancing.

The occurrence of multiple holes in wireless sensor networks poses many challenges in designing routing protocols. The traditional scheme is forwarding packets along the hole perimeters. However, this scheme leads to two serious problems: data concentration around the hole boundaries and routing path enlargement Recently, several approaches have been proposed to address these two problems, wherein a common idea is to form forbidden areas around the holes from which packets are kept to stay away. However, due to the static nature of the forbidden areas and routing paths, the existing protocols cannot solve these two problems thoroughly. In this paper, we propose a novel protocol for bypassing multiple holes in wireless sensor networks which can balance the traffic over the network while ensuring the constant stretch property of the routing path. Our main idea is to use elastic forbidden areas and dynamic routing paths. The theoretical analysis proves that the routing path stretch of the proposed protocol can be controlled to be as small as 1 + ϵ (for any predefined ϵ > 0), and the simulation experiments show that our protocol strongly outperforms state-of-the-art protocols in terms of load balancing.

There are two main requirements in dealing with coverage holes in wireless sensor networks (WSNs): locating the hole boundary and finding the locations to deploy new sensors for hole patching. The current protocols on finding the patching locations always require re-running the protocols from scratch many times. This constraint causes the time complexity and energy overhead to increase proportionally to the hole size. In this paper, we propose a lightweight protocol to determine coverage holes in wireless sensor network. Our protocol does not only can determine the exact hole boundary but also approximate the boundary by a simpler shape which can help to speed up the patching location finding process. The simulation experiments show that our protocol can reduce more than 56% of time complexity and save more than 46% of energy overhead in comparison with existing protocols.

The routing stretch in a sensor network with a large hole can be controlled if the network nodes have certain information about the hole location and shape, possibly by using some mechanism for hole boundary approximation and information dissemination. However, a finer approximation (for smaller stretch) leads to higher cost of dissemination, which could be unaffordable for the sensors. We provide an insightful analysis to this fundamental trade-off of the hole approximation problem, using a geometric model. Based on this we discuss an efficient approximation-dissemination scheme using a natural, heuristic approximation technique. For reasonable conditions (network is dense enough apart from the hole), we assure that the routing stretch is under a predefined threshold > 1 while the dissemination cost is also under tight control. Our initial experiment results also confirm that our approach is significantly more efficient and economical, compared to existing proposals.