Abstract: Unmanned aerial vehicle (UAV) communication is anticipated to be widely applied in the forthcoming fifth-generation wireless networks, due to its many advantages such as low cost, high mobility, and on-demand deployment. However, the broadcast and line-of-sight nature of air-to-ground wireless channels give rise to a new challenge on how to realize secure UAV communications with the destined nodes on the ground. This paper aims to tackle this challenge by applying the physical layer security technique. We consider both the downlink and uplink UAV communications with a ground node, namely, UAV-to-ground (U2G) and ground-to-UAV (G2U) communications, respectively, subject to a potential eavesdropper on the ground. In contrast to the existing literature on the wireless physical layer security only with the ground nodes at fixed or quasi-static locations, we exploit the high mobility of the UAV to proactively establish favorable and degraded channels for the legitimate and eavesdropping links, through its trajectory design. We formulate new problems to maximize the average secrecy rates of the U2G and G2U transmissions, by jointly optimizing the UAV’s trajectory, and the transmit power of the legitimate transmitter over a given flight period of the UAV. Although the formulated problems are non-convex, we propose iterative algorithms to solve them efficiently by applying the block coordinate descent and successive convex optimization methods. Specifically, both the transmit power and UAV trajectory are optimized, with the other being fixed in an alternating manner, until the algorithms converge. The simulation results show that the proposed algorithms can improve the secrecy rates for both U2G and G2U communications, as compared to other benchmark schemes without power control and/or trajectory optimization.

Abstract: We present network embedding algorithms that capture information about a node from the local distribution over node attributes around it, as observed over random walks following an approach similar to Skip-gram. Observations from neighborhoods of different sizes are either pooled (AE) or encoded distinctly in a multi-scale approach (MUSAE). Capturing attribute-neighborhood relationships over multiple scales is useful for a diverse range of applications, including latent feature identification across disconnected networks with similar attributes. We prove theoretically that matrices of node-feature pointwise mutual information are implicitly factorized by the embeddings. Experiments show that our algorithms are robust, computationally efficient and outperform comparable models on social networks and web graphs.

Abstract: This white paper describes the research and development needed over the next decade to realize “Cosmic Explorer,” the U.S. node of a future third-generation detector network that will be capable of observing and characterizing compact gravitational-wave sources to cosmological redshifts.

Abstract: Graph Convolutional Networks (GCNs) are an emerging type of neural network model on graphs which have achieved state-of-the-art performance in the task of node classification. However, recent studies show that GCNs are vulnerable to adversarial attacks, i.e. small deliberate perturbations in graph structures and node attributes, which poses great challenges for applying GCNs to real world applications. How to enhance the robustness of GCNs remains a critical open problem. To address this problem, we propose Robust GCN (RGCN), a novel model that "fortifies'' GCNs against adversarial attacks. Specifically, instead of representing nodes as vectors, our method adopts Gaussian distributions as the hidden representations of nodes in each convolutional layer. In this way, when the graph is attacked, our model can automatically absorb the effects of adversarial changes in the variances of the Gaussian distributions. Moreover, to remedy the propagation of adversarial attacks in GCNs, we propose a variance-based attention mechanism, i.e. assigning different weights to node neighborhoods according to their variances when performing convolutions. Extensive experimental results demonstrate that our proposed method can effectively improve the robustness of GCNs. On three benchmark graphs, our RGCN consistently shows a substantial gain in node classification accuracy compared with state-of-the-art GCNs against various adversarial attack strategies.

Abstract: This white paper describes the research and development needed over the next decade to realize "Cosmic Explorer," the U.S. node of a future third-generation detector network that will be capable of observing and characterizing compact gravitational-wave sources to cosmological redshifts.

Abstract: Learning node embeddings that capture a node's position within the broader graph structure is crucial for many prediction tasks on graphs. However, existing Graph Neural Network (GNN) architectures have limited power in capturing the position/location of a given node with respect to all other nodes of the graph. Here we propose Position-aware Graph Neural Networks (P-GNNs), a new class of GNNs for computing position-aware node embeddings. P-GNN first samples sets of anchor nodes, computes the distance of a given target node to each anchor-set,and then learns a non-linear distance-weighted aggregation scheme over the anchor-sets. This way P-GNNs can capture positions/locations of nodes with respect to the anchor nodes. P-GNNs have several advantages: they are inductive, scalable,and can incorporate node feature information. We apply P-GNNs to multiple prediction tasks including link prediction and community detection. We show that P-GNNs consistently outperform state of the art GNNs, with up to 66% improvement in terms of the ROC AUC score.

Abstract: Recently, graph neural networks have attracted great attention and achieved prominent performance in various research fields. Most of those algorithms have assumed pairwise relationships of objects of interest. However, in many real applications, the relationships between objects are in higher-order, beyond a pairwise formulation. To efficiently learn deep embeddings on the high-order graph-structured data, we introduce two end-to-end trainable operators to the family of graph neural networks, i.e., hypergraph convolution and hypergraph attention. Whilst hypergraph convolution defines the basic formulation of performing convolution on a hypergraph, hypergraph attention further enhances the capacity of representation learning by leveraging an attention module. With the two operators, a graph neural network is readily extended to a more flexible model and applied to diverse applications where non-pairwise relationships are observed. Extensive experimental results with semi-supervised node classification demonstrate the effectiveness of hypergraph convolution and hypergraph attention.

Abstract: Consensus is fundamental for distributed systems since it underpins key functionalities of such systems ranging from distributed information fusion, decision making, to decentralized control. In order to reach an agreement, existing consensus algorithms require each agent to exchange explicit state information with its neighbors. This leads to the disclosure of private state information, which is undesirable in cases where privacy is of concern. In this paper, we propose a novel approach for undirected networks, which can enable secure and privacy-preserving average consensus in a decentralized architecture in the absence of an aggregator or third party. By leveraging partial homomorphic cryptography to embed secrecy in pairwise interaction dynamics, our approach can guarantee convergence to the consensus value (subject to a quantization error) in a deterministic manner without disclosing a node's state to its neighbors. We provide a new privacy definition for dynamical systems, and give a new framework to rigorously prove that a node's privacy can be protected as long as it has at least one legitimate neighbor, which follows the consensus protocol faithfully without attempts to infer other nodes’ states. In addition to enabling resilience to passive attackers aiming to steal state information, the approach also allows easy incorporation of defending mechanisms against active attackers who try to alter the content of exchanged messages. Furthermore, in contrast to existing noise-injection-based privacy-preserving mechanisms that have to reconfigure the entire network when the topology or number of nodes varies, our approach is applicable to dynamic environments with time-varying coupling topologies. This secure and privacy-preserving approach is also applicable to weighted average consensus as well as maximum/minimum consensus under a new update rule. Numerical simulations and comparison with existing approaches confirm the theoretical results. Experimental results on a Raspberry-Pi board based microcontroller network are also presented to verify the effectiveness and efficiency of the approach.

Abstract: The smart vehicles construct Internet of Vehicle (IoV), which can execute various intelligent services. Although the computation capability of a vehicle is limited, multi-type of edge computing nodes provide heterogeneous resources for intelligent vehicular services. When offloading the complex service to the vehicular edge computing node, the decision for its destination should be considered according to numerous factors. This paper mostly formulate the offloading decision as a resource scheduling problem with single or multiple objective function and constraints, where some customized heuristics algorithms are explored. However, offloading multiple data dependence tasks in a complex service is a difficult decision, as an optimal solution must understand the resource requirement, the access network, the user mobility, and importantly the data dependence. Inspired by recent advances in machine learning, we propose a knowledge driven (KD) service offloading decision framework for IoV, which provides the optimal policy directly from the environment. We formulate the offloading decision for the multiple tasks as a long-term planning problem, and explore the recent deep reinforcement learning to obtain the optimal solution. It can scruple the future data dependence of the following tasks when making decision for a current task from the learned offloading knowledge. Moreover, the framework supports the pre-training at the powerful edge computing node and continually online learning when the vehicular service is executed, so that it can adapt the environment changes and can learn policy that are sensible in foresight. The simulation results show that KD service offloading decision converges quickly, adapts to different conditions, and outperforms a greedy offloading decision algorithm.

Abstract: A vehicular ad-hoc network (VANET) can improve the flow of traffic to facilitate intelligent transportation and to provide convenient information services, where the goal is to provide self-organizing data transmission capabilities for vehicles on the road to enable applications, such as assisted vehicle driving and safety warnings. VANETs are affected by issues such as identity validity and message reliability when vehicle nodes share data with other nodes. The method used to allow the vehicle nodes to upload sensor data to a trusted center for storage is susceptible to security risks, such as malicious tampering and data leakage. To address these security challenges, we propose a data security sharing and storage system based on the consortium blockchain (DSSCB). This digital signature technique based on the nature of bilinear pairing for elliptic curves is used to ensure the reliability and integrity when transmitting data to a node. The emerging consortium blockchain technology provides a decentralized, secure, and reliable database, which is maintained by the entire network node. In DSSCB, smart contracts are used to limit the triggering conditions for preselected nodes when transmitting and storing data and for allocating data coins to vehicles that participate in the contribution of data. The security analysis and performance evaluations demonstrated that our DSSCB solution is more secure and reliable in terms of data sharing and storage. Compared with the traditional blockchain system, the time required to confirm the data block was reduced by nearly six times and the transmission efficiency was improved by 83.33%.

Abstract: Graph matching refers to finding node correspondence between graphs, such that the corresponding node and edge's affinity can be maximized. In addition with its NP-completeness nature, another important challenge is effective modeling of the node-wise and structure-wise affinity across graphs and the resulting objective, to guide the matching procedure effectively finding the true matching against noises. To this end, this paper devises an end-to-end differentiable deep network pipeline to learn the affinity for graph matching. It involves a supervised permutation loss regarding with node correspondence to capture the combinatorial nature for graph matching. Meanwhile deep graph embedding models are adopted to parameterize both intra-graph and cross-graph affinity functions, instead of the traditional shallow and simple parametric forms e.g. a Gaussian kernel. The embedding can also effectively capture the higher-order structure beyond second-order edges. The permutation loss model is agnostic to the number of nodes, and the embedding model is shared among nodes such that the network allows for varying numbers of nodes in graphs for training and inference. Moreover, our network is class-agnostic with some generalization capability across different categories. All these features are welcomed for real-world applications. Experiments show its superiority against state-of-the-art graph matching learning methods.

Abstract: The shortage of spectrum resources has limited the development of Internet of Things (IoT). Fifth generation (5G) network can flexibly support a variety of devices and services, which makes it possible to combine 5G with IoT. In this paper, a novel multichannel IoT is proposed to dynamically share the spectrum with 5G communication, where an IoT node including transmitter and receiver is designed to perform 5G communication and IoT communication simultaneously. The subchannel sets allocated for 5G communication and IoT communication are defined by two complementary spectrum marker vectors, respectively. Two independent spectrum sequences are generated by calculating the inner products of spectrum marker vectors, presudo-random phases and power scaling vectors. Two time-domain fundamental modulation waveforms generated by the inverse fast Fourier transform of the spectrum sequences are used to modulate 5G data and IoT data, respectively. The receiver can detect the data using the same spectrum marker vectors as the transmitter. The BER performances of the system using binary modulation and cyclic code shift keying modulation in the cases of spectrum marker error and multiple access are analyzed, respectively. A subchannel and power optimization unit is formulated as a joint optimization problem, which seeks to maximize the 5G throughput under the constraints of minimal IoT throughput, maximal power, and maximal interference. An alternative optimization problem is proposed to maximize the IoT throughput while guaranteeing the minimal 5G throughput. A joint optimization algorithm based on Lagrange dual decomposition is proposed to achieve the optimal solution. Simulation results indicate that the proposed IoT can improve the 5G throughput significantly while the IoT throughput is guaranteed.

Abstract: Blockchain has shown a great potential in Internet of Things (IoT) ecosystems for establishing trust and consensus mechanisms without involvement of any third party. Understanding the relationship between communication and blockchain as well as the performance constraints posing on the counterparts can facilitate designing a dedicated blockchain-enabled IoT systems. In this paper, we establish an analytical model for the blockchain-enabled wireless IoT system. By considering spatio-temporal domain Poisson distribution, i.e., node geographical distribution in spatial domain and transaction arrival rate in time domain are both modeled as Poisson point process (PPP), we first derive the distribution of signal-to-interference-plus-noise ratio (SINR), blockchain transaction successful rate as well as overall throughput. Based on the system model and performance analysis, we design an algorithm to determine the optimal full function node deployment for blockchain system under the criterion of maximizing transaction throughput. Finally, the security performance is analyzed in the proposed networks with three typical attacks. Solutions such as physical layer security are presented and discussed to keep the system secure under these attacks. Numerical results validate the accuracy of our theoretical analysis and optimal node deployment algorithm.

Abstract: Home automation systems have attracted considerable attention with the advancement of communications technology. A smart home (SH) is an Internet of Things (IoT) application that utilizes the Internet to monitor and control appliances using a home automation system. Lack of IoT technology usage, unfriendly user interface, limited wireless transmission range, and high costs are the limitations of existing home automation systems. Therefore, this study presents a cost-effective and hybrid (local and remote) IoT-based home automation system with a user-friendly interface for smartphones and laptops. A prototype called IoT@HoMe is developed with an algorithm to enable the monitoring of home conditions and automate the control of home appliances over the Internet anytime and anywhere. This system utilizes a node microcontroller unit (NodeMCU) as a Wi-Fi-based gateway to connect different sensors and updates their data to Adafruit IO cloud server. The collected data from several sensors (radio-frequency identification, ultrasonic, temperature, humidity, gas, and motion sensors) can be accessed via If This Then That (IFTTT) on users' devices (smartphones and/or laptops) over the Internet regardless of their location. A set of relays is used to connect the NodeMCU to homes under controlled appliances. The designed system is structured in a portable manner as a control box that can be attached for monitoring and controlling a real house. The proposed IoT-based system for home automation can easily and efficiently control appliances over the Internet and support home safety with autonomous operation. IoT@HoMe is a low cost and reliable automation system that reduces energy consumption and can notably provide convenience, safety, and security for SH residents.

Abstract: Fog Computing (FC) is a flexible architecture to support distributed domain-specific applications with cloud-like quality of service. However, current FC still lacks the mobility support mechanism when facing many mobile users with diversified application quality requirements. Such mobility support mechanism can be critical such as in the industrial internet where human, products, and devices are moveable. To fill in such gaps, in this paper we propose novel container migration algorithms and architecture to support mobility tasks with various application requirements. Our algorithms are realized from three aspects: 1) We consider mobile application tasks can be hosted in a container of a corresponding fog node that can be migrated, taking the communication delay and computational power consumption into consideration; 2) We further model such container migration strategy as multiple dimensional Markov Decision Process (MDP) spaces. To effectively reduce the large MDP spaces, efficient deep reinforcement learning algorithms are devised to achieve fast decision-making and 3) We implement the model and algorithms as a container migration prototype system and test its feasibility and performance. Extensive experiments show that our strategy outperforms the existing baseline approaches 2.9, 48.5 and 58.4 percent on average in terms of delay, power consumption, and migration cost, respectively.

Abstract: Network embedding, which aims to represent network data in a low-dimensional space, has been commonly adopted for analyzing heterogeneous information networks (HIN). Although exiting HIN embedding methods have achieved performance improvement to some extent, they still face a few major weaknesses. Most importantly, they usually adopt negative sampling to randomly select nodes from the network, and they do not learn the underlying distribution for more robust embedding. Inspired by generative adversarial networks (GAN), we develop a novel framework HeGAN for HIN embedding, which trains both a discriminator and a generator in a minimax game. Compared to existing HIN embedding methods, our generator would learn the node distribution to generate better negative samples. Compared to GANs on homogeneous networks, our discriminator and generator are designed to be relation-aware in order to capture the rich semantics on HINs. Furthermore, towards more effective and efficient sampling, we propose a generalized generator, which samples "latent" nodes directly from a continuous distribution, not confined to the nodes in the original network as existing methods are. Finally, we conduct extensive experiments on four real-world datasets. Results show that we consistently and significantly outperform state-of-the-art baselines across all datasets and tasks.

Abstract: In this paper, we consider the problem of task offloading in a software-defined access network, where IoT devices are connected to fog computing nodes by multi-hop IoT access-points (APs). The proposed scheme considers the following aspects in a fog-computing-based IoT architecture: 1) optimal decision on local or remote task computation; 2) optimal fog node selection; and 3) optimal path selection for offloading. Accordingly, we formulate the multi-hop task offloading problem as an integer linear program (ILP). Since the feasible set is non-convex, we propose a greedy-heuristic-based approach to efficiently solve the problem. The greedy solution takes into account delay, energy consumption, multi-hop paths, and dynamic network conditions, such as link utilization and SDN rule-capacity. Experimental results show that the proposed scheme is capable of reducing the average delay and energy consumption by 12% and 21%, respectively, compared with the state of the art.

Abstract: Fog computing is a paradigm that extends the cloud to intermediate network devices with computational and storage capacities. This allows the execution of applications closer to edge devices and end-users by allocating services in those intermediate devices. The placement of those services has an influence on the performance of the fog architecture. We propose a fog computing simulator for analyzing the design and deployment of applications through customized and dynamical strategies. We model the relationships among deployed applications, network connections, and infrastructure characteristics through complex network theory, enabling the integration of topological measures in dynamic and customizable strategies, such as the placement of application modules, workload location, and path routing and scheduling of services. We present a comparative analysis of the efficiency and the convergence of results of our simulator with the most referenced one, iFogSim. To highlight the YAFS functionalities, we model three scenarios that, to the best of our knowledge, cannot be implemented with current fog simulators: dynamic allocation of new application modules, dynamic failures of network nodes, and user mobility along with the topology.

Abstract: Fifth generation wireless networks are expected to provide advanced capabilities and create new markets. Among the emerging markets, Internet of Things (IoT) use cases are standing out with the proliferation of a wide range of sensors that can be configured to continuously monitor and transmit data for intelligent processing and decision making. Devices in such scenarios are normally extremely energy-constrained and often exist in large numbers and can be located in hard-to-reach areas; the fact that necessitates the design and implementation of effective energy-aware data collection mechanisms. To this end, we propose the utilization of unmanned aerial vehicles (UAVs) to collect data in dense wireless sensor networks using projection-based compressive data gathering (CDG) as a novel solution methodology. CDG is utilized to aggregate data en-route from a large set of sensor nodes to selected projection nodes acting as cluster heads (CHs) in order to reduce the number of needed transmissions leading to notable energy savings and extended network lifetime. The UAV transfers the gathered data from the CHs to a remote sink node, e.g., a 5G cellular base station, which avoids the need for long range transmissions or multihop communications among the sensors. Our problem definition aims at clustering the sensors, constructing an optimized forwarding tree per cluster, and gathering the data from selected CH nodes based on projection-based CDG with minimized UAV trajectory distance. We formulate a joint optimization problem and divide it into four complementary subproblems to generate close-to-optimal results with lower complexity. Moreover, we propose a set of effective algorithms to generate solutions for relatively large-scale network scenarios. We demonstrate the superiority of the proposed approach and the designed algorithms via detailed performance results with analysis, comparisons, and insights.

Abstract: Domain Adaptation (DA) approaches achieved significant improvements in a wide range of machine learning and computer vision tasks (i.e., classification, detection, and segmentation). However, as far as we are aware, there are few methods yet to achieve domain adaptation directly on 3D point cloud data. The unique challenge of point cloud data lies in its abundant spatial geometric information, and the semantics of the whole object is contributed by including regional geometric structures. Specifically, most general-purpose DA methods that struggle for global feature alignment and ignore local geometric information are not suitable for 3D domain alignment. In this paper, we propose a novel 3D Domain Adaptation Network for point cloud data (PointDAN). PointDAN jointly aligns the global and local features in multi-level. For local alignment, we propose Self-Adaptive (SA) node module with an adjusted receptive field to model the discriminative local structures for aligning domains. To represent hierarchically scaled features, node-attention module is further introduced to weight the relationship of SA nodes across objects and domains. For global alignment, an adversarial-training strategy is employed to learn and align global features across domains. Since there is no common evaluation benchmark for 3D point cloud DA scenario, we build a general benchmark (i.e., PointDA-10) extracted from three popular 3D object/scene datasets (i.e., ModelNet, ShapeNet and ScanNet) for cross-domain 3D objects classification fashion. Extensive experiments on PointDA-10 illustrate the superiority of our model over the state-of-the-art general-purpose DA methods.

Abstract: Nowadays, deep neural networks (DNNs) have become the main instrument for machine learning tasks within a wide range of domains, including vision, NLP, and speech. Meanwhile, in an important case of heterogenous tabular data, the advantage of DNNs over shallow counterparts remains questionable. In particular, there is no sufficient evidence that deep learning machinery allows constructing methods that outperform gradient boosting decision trees (GBDT), which are often the top choice for tabular problems. In this paper, we introduce Neural Oblivious Decision Ensembles (NODE), a new deep learning architecture, designed to work with any tabular data. In a nutshell, the proposed NODE architecture generalizes ensembles of oblivious decision trees, but benefits from both end-to-end gradient-based optimization and the power of multi-layer hierarchical representation learning. With an extensive experimental comparison to the leading GBDT packages on a large number of tabular datasets, we demonstrate the advantage of the proposed NODE architecture, which outperforms the competitors on most of the tasks. We open-source the PyTorch implementation of NODE and believe that it will become a universal framework for machine learning on tabular data.

Abstract: Federated learning is a distributed machine learning approach for computing models over data collected by edge devices. Most importantly, the data itself is not collected centrally, but a master-worker architecture is applied where a master node performs aggregation and the edge devices are the workers, not unlike the parameter server approach. Gossip learning also assumes that the data remains at the edge devices, but it requires no aggregation server or any central component. In this empirical study, we present a thorough comparison of the two approaches. We examine the aggregated cost of machine learning in both cases, considering also a compression technique applicable in both approaches. We apply a real churn trace as well collected over mobile phones, and we also experiment with different distributions of the training data over the devices. Surprisingly, gossip learning actually outperforms federated learning in all the scenarios where the training data are distributed uniformly over the nodes, and it performs comparably to federated learning overall.

Abstract: This paper is concerned with the moving horizon estimation problem for a class of discrete time-delay systems under the Round-Robin (RR) protocol. The communication between the sensor nodes and the remote state estimator is implemented via a shared network, where only one sensor node is permitted to transmit data at each time instant for the purpose of preventing data collisions. The RR protocol is utilized to orchestrate the transmission order of sensor nodes, under which the selected node obtaining access to the network could be modeled by a periodic function. A lifting technology is introduced to reformulate the system model into a linear system without delays. The aim of the addressed problem is to develop a moving horizon estimator such that the estimation error is ultimately bounded. A sufficient condition is established to ensure the ultimate boundedness in terms of a matrix inequality. Within the established theoretical framework, two optimization problems are proposed to calculate the corresponding estimator parameters according to two different performance requirements (e.g., the smallest ultimate bound and the fastest decay rate). Finally, simulation examples are given to illustrate the effectiveness of the estimator design scheme.

Abstract: To meet the requirements of long range, a small amount of data transmission, low power, and low cost of the Internet of Things (IoT) in actual applications, a low-power wide-area network information monitoring approach based on NB-IoT and LoRa is proposed in this paper. This approach adopts a communication mode that contains a main node and multiple subnodes to adapt to the needs of large-scale information monitoring. Among them, the design of the main node utilizes NB-IoT communication technology, LoRa communication technology, and least recently used algorithm. The design of subnode utilizes LoRa communication technology, sensor technology, and optical-electric conversion technology. Finally, we design and implement a cloud service and computing system, using the Tencent Cloud server. The advantage of this approach is that the combination of NB-IoT and LoRa not only improves transmission distance but also reduces the operating costs of the WAN information monitoring system. In addition, using optical-electric conversion technology, the system can be self-powered. Combined with the principle of electric circuits, the power consumption of the system can also be reduced. Finally, by conducting a system test, LoRa communication distance experiment, NB-IoT communication, and other experiments, the communication distance in a complex environment is up to 1.6 km, the minimum working current is 2 mA, and the system communication packet loss rate is approximately 3%. The system runs stably and the collected data are accurate. The results show that the proposed approach can meet the requirements of data collection, transmission, storage, and calculation in a wide area.

Abstract: One-Shot Neural Architecture Search (NAS) is a promising method to significantly reduce search time without any separate training. It can be treated as a Network Compression problem on the architecture parameters from an over-parameterized network. However, there are two issues associated with most one-shot NAS methods. First, dependencies between a node and its predecessors and successors are often disregarded which result in improper treatment over zero operations. Second, architecture parameters pruning based on their magnitude is questionable. In this paper, we employ the classic Bayesian learning approach to alleviate these two issues by modeling architecture parameters using hierarchical automatic relevance determination (HARD) priors. Unlike other NAS methods, we train the over-parameterized network for only one epoch then update the architecture. Impressively, this enabled us to find the architecture on CIFAR-10 within only 0.2 GPU days using a single GPU. Competitive performance can be also achieved by transferring to ImageNet. As a byproduct, our approach can be applied directly to compress convolutional neural networks by enforcing structural sparsity which achieves extremely sparse networks without accuracy deterioration.

Abstract: This paper investigated the multiple unmanned aerial vehicle (UAV) relays' assisted network in the Internet of Things (IoT) systems enhanced with energy harvesting in order to overcome the large-scale fading between source and sink as well as achieve the green cooperative communications, where time switch (TS) and power splitting (PS) strategies were typically applied for UAV relays to implement energy harvesting transmission, which was also selected via signal to noise ratio (SNR) maximization criterion so that the terminal node can obtain the optimal signal. Meanwhile, it was worth noting that the terminal node may be disturbed by aggregated interference caused by dense network signaling interaction in the future 5G/B5G systems. Therefore, after TS and PS protocols designing and utilizing, the closed-form expressions of outage probability and bit error rate (BER) for UAV relay assisted IoT systems suffered from aggregated interference were derived in detail. In addition, the throughput and delay limited state of UAV relay assisted transmission were also analyzed thoroughly. The derivations and analysis results showed that the proposed multi-parameter joint optimization of transmitting power, scaling factor, and UAV relay selection could effectively improve the system throughput and reduce the system outage probability and BER. The simulation experiments verified the effectiveness of the proposed schemes and the correctness of theoretical analysis.