Sequential federated learning (SFL) trains models collaboratively across clients in a chain manner. This order shows communication efficiency compared to traditional FL in a parallel manner with a star topology. However, SFL can fail to produce stable training results when clients have significant statistical heterogeneity among their local data distributions. To address these challenges, we propose a novel element-wise model decoupling framework namedSeqFedEDTthat accelerates SFL training by separating model parameters of each client into a shared subset for global knowledge collaboration and a personalized subset for migrating data heterogeneity. We explore three types of parameter contribution scoring metrics based on gradient, Fisher information, and parameter importance (PI) for personalized parameter selection. In addition, we propose a quantile-based thresholding mechanism to separate shared and personalized subsets and explore the best performance quantile selection in numerical studies. Extensive experiments demonstrate thatSeqFedEDToutperforms eight state-of-the-art methods across diverse datasets and heterogeneity scenarios. All code and results are available athttps://github.com/tian0920/SeqFedEDT.

Sparse tensor computing is widely used in deep learning and scientific computing, but its irregularity challenges GPU/NPU acceleration. FPGAs, with their reconfigurability, are well-suited for sparse workloads. However, current point-wise, manual design methodologies significantly limit the performance potential of reconfigurable hardware across diverse sparse scenarios. This paper proposes a compiler framework that automatically generates high-performance sparse accelerators on FPGA. It includes: (1) a schedule-primitive-based DSL serving as sparse accelerator design specification, covering a wide range of design concerns; (2) a two-stage mapping mechanism based on sparse meta-operations, enabling flexible mapping from arbitrary sparse dataflow to hardware microarchitectures; and (3) a heuristic design space exploration strategy guided by workload characteristics.

Enabling on-device intelligence that supports AI inference, has become a main trend in mobile application development. To enhance intelligence on resource-limited smartphones in mobile cellular networks, edge computing offloads tasks to powerful servers. However, current AI inference models exhibit significant heterogeneity across various platforms, posing challenges to edge offloading implementation. The complexity of managing these heterogeneous models is exacerbated by the super-large scale of smartphones and the need for realtime information exchange. These challenges, stemming from prohibitive communication overheads, render current offloading approaches economically undesirable. This paper proposes LOSA, a decentralized task-offloading scheme for edge-assisted AI inference systems with three main components. First, a performance estimator to measure the gains in offloading diverse AI model execution to the edge. Second, a workload forecaster that predicts the future workload based on historical traces. Last, a scheduler determines the optimal place to execute the model by solving a stochastic optimization problem with local model performance and workload prediction. Simulation results show how LOSA outperforms the current solutions while maintaining low communication overhead.

Large language models (LLMs) are expected to offer structured Markdown responses for the sake of readability in web chatbots (e.g., ChatGPT). Although there are a myriad of metrics to evaluate LLMs, they fail to evaluate the readability from the view of output content structure. To this end, we focus on an overlooked yet important metric --- Markdown Awareness, which directly impacts the readability and structure of the content generated by these language models. In this paper, we introduce MDEval, a comprehensive benchmark to assess Markdown Awareness for LLMs, by constructing a dataset with 20K instances covering 10 subjects in English and Chinese. Unlike traditional model-based evaluations, MDEval provides excellent interpretability by combining model-based generation tasks and statistical methods. Our results demonstrate that MDEval achieves a Spearman correlation of 0.791 and an accuracy of 84.1% with human, outperforming existing methods by a large margin. Extensive experimental results also show that through fine-tuning over our proposed dataset, less performant open-source models are able to achieve comparable performance to GPT-4o in terms of Markdown Awareness. To ensure reproducibility and transparency, MDEval is open sourced at https://github.com/SWUFE-DB-Group/MDEval-Benchmark.

Federated learning, as a promising distributed learning paradigm, enables collaborative training of a global model across multiple network edge clients without the need for central data collecting. However, the heterogeneity of edge data distribution drags the model towards the local minima, which can be distant from the global optimum. Such heterogeneity often leads to slow convergence and substantial communication overhead. To address these issues, we propose a novel federated learning framework calledFedCMD, a model decoupling tailored to the Cloud-edge supported federated learning that separates deep neural networks into a body for capturing shared representations in Cloud and a personalized head for migrating data heterogeneity. Our motivation is that, by the deep investigation of the performance of selecting different neural network layers as the personalized head, we found rigidly assigning the last layer as the personalized head in current studies is not always optimal. Instead, it is necessary to dynamically select the personalized layer that maximizes the training performance by taking the representation difference between neighbor layers into account. To find the optimal personalized layer, we utilize the low-dimensional representation of each layer to contrast feature distribution transfer and introduce a Wasserstein-based layer selection method, aimed at identifying the best-match layer for personalization. Additionally, a weighted global aggregation algorithm is proposed based on the selected personalized layer for the practical application ofFedCMD. Extensive experiments on ten benchmarks demonstrate the efficiency and superior performance of our solution compared with nine state-of-the-art solutions. All code and results are available athttps://github.com/elegy112138/FedCMD.

Unmanned aerial vehicles (UAV) are widely used for edge computing in poor infrastructure scenarios due to their deployment flexibility and mobility. In UAV-assisted edge computing systems, multiple UAVs can cooperate with the cloud to provide superior computing capability for diverse innovative services. However, many service-related computational tasks may fail due to the unreliability of UAVs and wireless transmission channels. Diverse solutions were proposed, but most of them employ timedriven strategies which introduce unwanted decision waiting delays. To address this problem, this paper focuses on a taskdriven reliability-aware cooperative offloading problem in UAV-assisted edge-enhanced networks. The issue is formulated as an optimization problem which jointly optimizes UAV trajectories, offloading decisions, and transmission power, aiming to maximize the long-term average task success rate. Considering the discrete-continuous hybrid action space of the problem, a dependenceaware latent-space representation algorithm is proposed to represent discrete-continuous hybrid actions. Furthermore, we design a novel deep reinforcement learning scheme by combining the representation algorithm and a twin delayed deep deterministic policy gradient algorithm. We compared our proposed algorithm with four alternative solutions via simulations and a realistic Kubernetes testbed-based setup. The test results show how our scheme outperforms the other methods, ensuring significant improvements in terms of task success rate.

Temporal graph representation learning has drawn considerable attention in recent years. Most existing works mainly focus on modeling local structural dependencies of temporal graphs. However, underestimating the inherent global structural role information in many real-world temporal graphs inevitably leads to sub-optimal graph representations. To overcome this shortcoming, we propose a novel Role-based Temporal Graph Convolution Network (RTGCN) that fully leverages the global structural role information in temporal graphs. Specifically, RTGCN can effectively capture the static global structural roles by using hypergraph convolution neural networks. To capture the evolution of nodes' structural roles, we further design structural role-based gated recurrent units. Finally, we integrate structural role proximity in our objective function to preserve global structural similarity, further promoting temporal graph representation learning. Experimental results on multiple real-world datasets demonstrate that RTGCN consistently outperforms state-of-the-art temporal graph representation learning methods by significant margins in various temporal link prediction and node classification tasks. Specifically, RTGCN achieves AUC improvement of up to 5.1% for link prediction and F1 improvement of up to 6.2% for new link prediction. In addition, RTGCN achieves AUC improvement up to 4.6% for node classification and 2.7% for structural role classification.

Abstract Whole brain segmentation from magnetic resonance images (MRI) is crucial in diagnosing brain diseases and analyzing neuroimaging data. Despite advances through deep learning, challenges such as uneven gray distribution and the presence of artifacts still present hurdles in medical image processing. These limitations are often a result of insufficient spatial contextual information and lack of attention to important regions within existing models. To address these issues, this paper presents SSA‐UNet (Squeeze‐and‐Excitation and Self‐Attention UNet), a uniquely designed deep convolutional neural network that integrates spatial constraints by converting three consecutive 2D MRI slices into a single 2.5D image. This facilitates capturing inter‐slice dependencies effectively. Additionally, the newly formulated SSA block, which sequentially incorporates channel attention and Self‐Attention mechanisms, is placed before the decoders in the conventional U‐Net architecture. This enables the network to automatically weight different feature maps and focus more effectively on regions requiring precise segmentation. Rigorous evaluations on LPBA40 and IBSR18 datasets substantiate the remarkable improvements in accuracy and stability achieved by SSA‐UNet. Results indicate Dice coefficients of 98.38% and 97.47%, specificity of 99.69% and 99.57%, and sensitivity of 98.5% and 97.98% for the respective datasets. Compared to other existing models, SSA‐UNet shows significant improvements on both the LPBA40 and IBSR18 datasets. On the LPBA40 dataset, SSA‐UNet's Dice coefficient improved by 0.33% compared to the sub‐optimal model, while on the IBSR18 dataset, the improvement reached 1.78%. These empirical findings demonstrate SSA‐UNet's heightened capability in addressing the long‐standing challenges in MRI‐based whole‐brain segmentation.

The exceptionally high bandwidth requirement for delivering high-quality live 360$^\circ$video poses a significant challenge to current network capacity. Mitigating such bandwidth starvation necessitates accurate field-of-view (FoV) prediction to focus limited resources on the viewer's area of interest. However, FoV prediction for live 360$^\circ$streaming can be complex due to the time-sensitive nature of live content and the limited knowledge available for model training. Our paper introduces a novel framework,CoLive, for predicting the FoV in 360$^\circ$live streaming.CoLiveaccelerates FoV prediction by offloading model training from viewers to the edge and migrating saliency feature detection to the server side. Observations on user clustering of viewing behaviors further motivate us to propose a novel dynamic clustered learning algorithm. The algorithm dynamically groups users according to their model update gradients and enables them to train a shared model that better suits their viewing preferences. We conduct extensive experiments on the public 360$^\circ$video datasets and demonstrate thatCoLiveoutperforms state-of-the-art solutions in terms of prediction performance and bandwidth savings.

Stock Movement Prediction (SMP) aims at predicting listed companies' stock future price trend, which is a challenging task due to the volatile nature of financial markets. Recent financial studies show that the momentum spillover effect plays a significant role in stock fluctuation. However, previous studies typically only learn the simple connection information among related companies, which inevitably fail to model complex relations of listed companies in real financial market. To address this issue, we first construct a more comprehensive Market Knowledge Graph (MKG) which contains bi-typed entities including listed companies and their associated executives, and hybrid-relations including the explicit relations and implicit relations. Afterward, we proposeDanSmp, a novel Dual Attention Networks to learn the momentum spillover signals based upon the constructed MKG for stock prediction. The empirical experiments on our constructed datasets against nine SOTA baselines demonstrate that the proposedDanSmpis capable of improving stock prediction with the constructed MKG.

Bi-typed multi-relational heterogeneous graph (BMHG) is one of the most common graphs in practice, for example, academic networks, e-commerce user behavior graph and enterprise knowledge graph. It is a critical and challenge problem on how to learn the numerical representation for each node to characterize subtle structures. However, most previous studies treat all node relations in BMHG as the same class of relation without distinguishing the different characteristics between the intra-type relations and inter-type relations of the bi-typed nodes, causing the loss of significant structure information. To address this issue, we propose a novelDualHierarchicalAttentionNetworks (DHAN) based on the bi-typed multi-relational heterogeneous graphs to learn comprehensive node representations with the intra-type and inter-type attention-based encoder under a hierarchical mechanism. Specifically, the former encoder aggregates information from the same type of nodes, while the latter aggregates node representations from its different types of neighbors. Moreover, to sufficiently model node multi-relational information in BMHG, we adopt a newly proposed hierarchical mechanism. By doing so, the proposed dual hierarchical attention operations enable our model to fully capture the complex structures of the bi-typed multi-relational heterogeneous graphs. Experimental results on various tasks against the state-of-the-arts sufficiently confirm the capability of DHAN in learning node representations on the BMHGs.

The latest evolution of wireless communications enables user access rich Virtual Reality (VR) services via the Internet, including while on the move. However, providing a premium immersive experience for massive number of concurrent users with various device configurations is a significant challenge due to the ultra-high data rate and ultra-low delay requirements of live VR services. This paper introduces an innovative multi-user cost-efficient crowd-assisted delivery and computing (MEC-DC) framework, which leverages mobile edge computing and end-user resources to support high performance VR content delivery over 5G-and-beyond heterogeneous networks (5G-HetNets). The proposed MEC-DC framework is based on three main solutions. First is a novel buffer-nadir-based multicast (BNM) mechanism for VR transmissions over 5G-HetNets. BNM ensures smooth and synchronized user viewing experience by maximizing the average playback buffer-nadir of all participants with stochastic optimization. Second and third are practical distributed algorithms: the cost-efficient multicast-aware transcoding offloading (MATO) and crowd-assisted delivery algorithm (CAD) which optimize jointly multicast delivery and video transcoding. The algorithms optimality and complexity were investigated. The proposed MATO-CAD solution was evaluated with real datasets, trace-driven numerical simulations, and prototype-based experiments. The trace-driven experimental results showed how the proposed solution provides 18% throughput improvement, lowest delay and best playback freeze ratio in comparison with three other state-of-the-art solutions.

Providing premium panoramic livecast services to worldwide viewers considering their ultra-high data rate and delay-sensitivity is a significant challenge in the current network delivery environment. Therefore, it is important to design an efficient way of improving viewer quality of experience while conserving bandwidth resources. In this context, this paper introduces a novel cost-efficient federated transmission framework calledFedLiveand a set of algorithms to support it. First a gradient-based clustering method is proposed to group the geo-distributed viewers with similar viewing behavior into content delivery alliances by exploiting the geometric properties of the gradient loss. Next, aReinforcedVariationalInference (RVI) structure-based approach is proposed to assist with the collaborative training of the viewer field of view (FoV) prediction model while also accelerating the tile delivery process. A novel prediction-based asynchronous delivery algorithm is designed in which both the high accuracy FoV prediction and efficient live 360$^\circ$video transmission are achieved in a decentralized manner. FedLive was implemented for testing and an open source code is made available. Finally, the proposed solution was evaluated against a benchmark and three alternative state-of-the-art solutions using a real-world dataset. The experimental results show that our approach provides the highest prediction accuracy, better service performance, and saves bandwidth when compared with the other solutions.

The ever-increasing demand for bandwidth resources when delivering premium quality 360° video challenges the current network capacity. To alleviate such bandwidth pressure, it is imperative to predict the viewport via observing the content visual feature and historical viewing behaviors, which thereby allows the system to concentrate the limited resource on viewer's region of interest in 360° content. However, enabling accurate viewport prediction for 360° live streaming is non-trivial given the time-sensitive of live content and shortage of pre-knowledge on the visual features and viewing behaviors. In this paper, we propose CoLive, an edge-assisted online viewport prediction framework. CoLive incorporates edge computing to offload the prediction model training from viewers and migrates the saliency feature detection to the server side for reducing the processing delay. Viewers can also collaboratively train a central predicting model via sharing their loss gradients. This central model, together with the saliency feature detection, further prompts accuracy prediction and learning acceleration, especially for new incoming viewers. A series of experiments on the public 360° video dataset show how our solution achieves better performance compared with state-of-the-art solutions.

Recent years have witnessed an outstanding increase in popularity of Crowdsourced Livecast Services (CLS), which is the latest trend in social media. In CLS, transcoding enormous video contents from massive broadcasters and providing high-quality CLS for global viewers with heterogeneous devices are computation-intensive as well as time-consuming. There are some schemes that design viewer-assisted transcoding scheme, but it is challenging to achieve an efficient and fair task assignment due to the dynamic of computing and communication resources. This paper introduces a viewer-assisted CLS framework and focuses on proposing an innovative fairness-guaranteed task assignment scheme, which is a key challenge in this context. Considering the dynamic nature of viewers’ computing and communication resources and stability, a dynamic programming problem with fairness and QoS constraints is formulated. To solve the problem, we devise a Fair Bandit (FB) algorithm based on the Combinatorial Multi-Armed Bandit (CMAB). Finally, the effectiveness of proposed scheme is demonstrated by trace-driven simulations.

Intensive video transcoding and data transmission are the most crucial tasks for large-scale Crowd-sourced Livecast Services (CLS). However, there exists no versatile model for joint optimization of computing resources (e.g., CPU) and transmission resources (e.g., bandwidth) in CLS systems, making maintaining the balance between saving resources and improving user viewing experience very challenging. In this paper, we first propose a novel universal model, called Augmented Graph Model (AGM), which converts the above joint optimization into a multi-hop routing problem. This model provides a new perspective for the analysis of resource allocation in CLS, as well as opens new avenues for problem-solving. Further, we design a decentralized Networked Multi-Agent Reinforcement Learning (MARL) approach and propose an actor-critic algorithm, allowing network nodes (agents) to distributively solve the multi-hop routing problem using AGM in a fully cooperative manner. By leveraging the computing resource of massive nodes efficiently, this approach has good scalability and can be employed in large-scale CLS. To the best of our knowledge, this work is the first attempt to apply networked MARL on CLS. Finally, we use the centralized (single-agent) RL algorithm as a benchmark to evaluate the numerical performance of our solution in a large-scale simulation. Additionally, experimental results based on a prototype system show that our solution is superior in saving resources and service performance to two alternative state-of-the-art solutions.

Nowadays, amateur broadcasters can massively generate video contents and stream them across the Internet. For this reason, crowdsourced livecast services (CLS) are attracting millions of users around the world. To provide a smooth and high-quality playback experience to viewers with diversified device configurations in dynamic network conditions, CLS providers have to find a way to deploy cost-effective transcoding operations by distributing the computation-intensive workload among Cloud, Edge, and Crowd. In addition, it is necessary to control transcoded streams from million broadcasters to worldwide viewers. To address these challenges, we propose a novel stochastic approach that jointly optimizes the usage of transmission resources (e.g., bandwidth), and transcoding resources (e.g., CPU) in CLS systems that leverage the cooperation of Cloud, Edge, and Crowd technologies. In particular, we first design an augmented queue structure that can jointly capture the dynamic features of data transmission and online transcoding, based on the virtual queue technology. Then, we formulate a joint resource allocation problem, using stochastic optimization arguments, and devise an Accelerated Gradient Optimization (AGO) algorithm to solve the optimization problem in a scalable way. Moreover, we provide four main theoretical results that characterize the algorithm’s steady-state queue-length, optimality, and fast-convergence. By conducting both numerical simulations and system-level evaluations based on our prototype, we demonstrate that our solution provides lower system costs and higher QoE performance against state-of-the-art solutions.

By exploiting the massive data generated from the numerous interconnected machines and control systems, industrial Internet-of-Things (IIoT) provides unprecedented opportunities for facilitating the intelligence and smartness of manufacturing. Timely processing the large-scaled IIoT data by the conventional computation framework, such as Cloud computing, however, is nontrivial due to its costly resource usage, intolerable delay, and unbearable backbone pressures. By leveraging the idle resources of smart objects at the edge, mobile device cloud (MDC) becomes promising for the IIoT data analysis, thanks to the flexible resource provision and nearby task offloading. However, MDC workers are mostly human-carried devices with large scale, high dynamic resource provision, and untruthful behaviors, which pose significant challenges on MDC task allocation. In this article, we propose a blockchain-based decentralized and truthful framework for MDC (BC-MDC). BC-MDC enables the decentralization and prevents dishonesty by incorporating a plasma-based blockchain into the MDC. We design four smart contracts for distributedly managing the worker registration, task posting/allocation, rewarding, and penalizing. Furthermore, MDC task allocation is formulated as a stochastic optimization problem that jointly minimizes the long-term processing cost and risk of task failing. We also design a truthful reward/penalty algorithm that stimulates workers to provide resources and enforce them to keep the promise as well. Collaborated by the extensive simulation tests, we show how our proposed scheme achieves low cost on usage and high truthfulness and outperforms state-of-the-art solutions.

Information-centric networking (ICN) is of high interest to the Internet of Things (IoT) community, since the dissemination of massive data continuously produced by IoT devices can be easily handled by ICN’s data naming scheme and inherent multipath delivery. Providing optimal multipath-oriented transmission control is crucial for ICN-IoT data delivery, but yet remains challenging because of the randomness of request arrival, dynamic link condition, and on-path caching. More prominently, the resource limitation and scalability issues in IoT require the control scheme to be lightweight and distributed. In this paper, we propose a distributed stochastic optimization framework for multipath transmission control in ICN-IoT. The transmission control, including request scheduling and data rate regulation, is formulated as a stochastic concave optimization problem, which aims to accommodate the randomness, unpredictability, and multipath delivery of ICN-IoT and maximize the overall throughput. This problem is linearly separated into two subproblems: 1) a request scheduling problem and 2) a data rate control problem, which can be individually solved per time slot. A distributed alternating descent method (DADM) is designed to optimally control the transmission by solving the aforementioned problems at client sides. DADM enables each client to sequentially update the request schedule and rate regulation via communicating the links and providers they use, which asymptotically converges to optimality while allowing low-complexity and decentralized implementation. Validated by simulations, our DADM significantly improves throughput, delay reduction, and energy efficiency, in comparison with other state-of-the-art solutions.

The ever fast growing mobile social video traffic has motivated the urgent requirement of alleviating backbone pressures while ensuring the user-quality experience. Mobile video prefetching previously caches the future accessed videos at the edge, which has become a promising solution for traffic offloading and delay reduction. However, providing high performance prefetching still remains problematic in the presence of high dynamic mobile users' viewing behaviors and consecutive generated video content. Besides, given the fact that making prefetching decision requires viewing history that is sensitive, the increasing privacy issues should also be considered. In this paper, we propose a differential privacy oriented distributed online learning method for mobile social video prefetching (DPDL-SVP). Through a large-scale data analysis based on one of the most popular online social network sites, WeiBo.cn, we reveal that users' viewing behaviors have strong a relation with video preference, content popularity, and social interactions. We then formulate the prefetching problem as an online convex optimization based on these three factors. Furthermore, the problem is divided into two subproblems, and we implement a distributed algorithm separately to solve them with differential privacy. The performance bound of the proposed online algorithms is also theoretically proved. We conduct a series simulation based on real viewing traces to evaluate the performance of DPDL-SVP. Evaluation results show how our proposed algorithms achieve superior performance in terms of the prediction accuracy, delay reduction, and scalability.

The manyfold capacity magnification promised by dense 5G networks will make possible the provisioning of broadband multimedia services, including virtual reality, augmented reality, and mobile immersive video, to name a few. These new applications will coexist with classic ones and contribute to the exponential growth of multimedia services in mobile networks. At the same time, the different requirements of past and old services pose new challenges to the effective usage of 5G resources. In response to these challenges, a novel Stochastic Optimization framework for Green Multimedia Services named SOGMS is proposed herein that targets the maximization of system throughput and the minimization of energy consumption in data delivery. In particular, Lyapunov optimization is leveraged to face this optimization objective, which is formulated and decomposed into three tractable subproblems. For each subproblem, a distinct algorithm is conceived, namely quality of experience--based admission control, cooperative resource allocation, and multimedia services scheduling. Finally, extensive simulations are carried out to evaluate the proposed method against state-of-art solutions in dense 5G networks.

As one of the key technologies for future 5G, Device- to-Device communications (D2D) offloads traffic to local by enabling mobile equipment directly communicating with each other, which perfectly supporting distributed applications and IoT scenarios. Integrating Information-centric networking (ICN) with D2D is becoming an attractive trend because of the superior advantages of inherent support of caching and name-based routing. Nevertheless, efficient caching in ICN D2D still remain problematic due to the low utilization of caching space and multicast feature of wireless scenarios. In this paper, we propose a novel optimal coded content caching mechanism for ICN-based 5G D2D. We first building a fluid-based model to describe how the roles of mobile nodes evolve with the user behaviors and caching strategy. We then accordingly formulate the coded caching problem as an optimization problem, which mainly considers the tradeoff between delivery latency and energy consumption. The existence of optimal solutions is proved theoretically. We further propose a Learn Tree- based Code Content (LTCC) mechanism to cluster the contents for content coding selection and an Optimal Coded Content Caching (O3C) algorithm to solve coded content caching problem. Finally, we conduct massive simulation tests to validate the performance of the proposed algorithm against the state-of-art solutions.

The recent fast development of wireless communications and smart devices has opened the avenue to supporting high quality video streaming services in vehicular networks. This growing trend towards enhanced video services and the inefficient content distribution of conventional IP networks have motivated the researchers to propose new Internet architectures that are more efficient for content distribution in general and in vehicular networks in particular. Information-centric networking (ICN) shifts the network paradigm from host centric to content centric, providing effective content distribution by named-based routing and in-network caching, which becomes a promising solution for sharing video streaming among vehicles. In this paper, we present a novel Family-Aware Pricing Strategy (FAPS) to accelerate video streaming dissemination over Information-Centric Vehicular Networks (ICVNs). We first classify the mobile users into multiple families by investigating user similar behaviors. Based on the family, an efficient video sharing scheme is proposed to support near end video fetching. In addition, a pricing-based video caching policy is also proposed to accurately optimize caching distributions. Simulation results show how our proposed strategy achieves better performance than other state-of-art solutions in terms of caching hit ratio, searching delay, freeze times and control overhead.

Device-to-Device (D2D) communications are a prominent feature of 5G systems, introduced to provide a native support to distributed services in mobile environments. D2D technologies enable straight interactions between mobile terminals without a compulsory involvement of base stations. In this manuscript, we study and propose an optimized caching strategy to content distribution on top of D2D technology, based on Information Centric Networking (ICN) principles. The rationale is that ICN architectures can provide seamless support to mobile services and decouple contents from node identifiers, thus providing a promising match with D2D requirements. To this end, a novel fluid-based model in proposed hereby that catches the interplay between ICN functionalities, D2D requirements, and 5G specifications. Then, based on this model, an optimal content replication problem is formulated, encompassing caching overhead and system load. Additionally, this problem is thoroughly analyzed to prove that it has an optimal solution with time threshold form. A practical algorithm ς*-OCP is further proposed in order to implement the optimal caching control in realistic environments. Finally, a massive simulation campaign is carried out to test the proposed algorithm in comparison to state-of-the-art solutions.