arXiv subjects

Yan Zhang

Publications and source records attributed to Yan Zhang.

At least 541 records · Page 30Linked to original sources

arXiv2020

Edge Intelligence Empowered UAVs for Automated Wind Farm Monitoring in Smart Grids

With the exploitation of wind power, more turbines will be deployed at remote areas possibly with harsh working conditions (e.g., offshore wind farm). The adverse working environment may lead to massive operating and maintenance costs of turbines. Deploying unmanned aerial vehicles (UAVs) for turbine inspection is considered as a viable alternative to manual inspections. An important objective of automated UAV inspection is to minimize the flight time of the UAVs to inspect all the turbines. A first contribution of this paper is thus formulating an optimization problem to compute the optimal routes for turbine inspection satisfying the above goal. On the other hand, the limited computational capability on UAVs can be used to increase the power generation of wind turbine. Power generation from the turbines can be optimized by controlling the yaw angle of the turbines. Forecasting wind conditions such as wind speed and wind direction is crucial for solving both optimization problems. Therefore, UAVs can utilize their limited computational capability to perform wind forecasting. In this way, UAVs form edge intelligence in offshore wind farm. With the forecasted wind conditions, we design two algorithms to solve the formulated problems, and then evaluate the proposed methods with realworld data. The results reveal that the proposed methods offer an improvement of 44% of the power generation from the turbine compared to hour-ahead forecasting and 25% reduction of the flight time of the UAVs compared to the chosen baseline method.

eess.SY↗

arXiv2020

Time-Based Roofline for Deep Learning Performance Analysis

Deep learning applications are usually very compute-intensive and require a long run time for training and inference. This has been tackled by researchers from both hardware and software sides, and in this paper, we propose a Roofline-based approach to performance analysis to facilitate the optimization of these applications. This approach is an extension of the Roofline model widely used in traditional high-performance computing applications, and it incorporates both compute/bandwidth complexity and run time in its formulae to provide insights into deep learning-specific characteristics. We take two sets of representative kernels, 2D convolution and long short-term memory, to validate and demonstrate the use of this new approach, and investigate how arithmetic intensity, cache locality, auto-tuning, kernel launch overhead, and Tensor Core usage can affect performance. Compared to the common ad-hoc approach, this study helps form a more systematic way to analyze code performance and identify optimization opportunities for deep learning applications.

cs.DC↗

arXiv2020

A class of two or three weights linear codes and their complete weight enumerators

In the past few years, linear codes with few weights and their weight analysis have been widely studied. In this paper, we further investigate a class of two-weight or three-weight linear codes from defining sets and determine their weight and complete weight enumerators by application of the theory of quadratic forms and some special Weil sums over finite fields. Some punctured codes of the discussed linear codes are optimal or almost optimal with respect to the Griesmer bound. This paper generalizes some results in \cite{ZhuXu2017,Jian2019}.

cs.IT↗

arXiv2020

Attention Guided Graph Convolutional Networks for Relation Extraction

Dependency trees convey rich structural information that is proven useful for extracting relations among entities in text. However, how to effectively make use of relevant information while ignoring irrelevant information from the dependency trees remains a challenging research question. Existing approaches employing rule based hard-pruning strategies for selecting relevant partial dependency structures may not always yield optimal results. In this work, we propose Attention Guided Graph Convolutional Networks (AGGCNs), a novel model which directly takes full dependency trees as inputs. Our model can be understood as a soft-pruning approach that automatically learns how to selectively attend to the relevant sub-structures useful for the relation extraction task. Extensive results on various tasks including cross-sentence n-ary relation extraction and large-scale sentence-level relation extraction show that our model is able to better leverage the structural information of the full dependency trees, giving significantly better results than previous approaches.

cs.CL↗

arXiv2020

5G Technologies Based Remote E-Health: Architecture, Applications, and Solutions

Currently, many countries are facing the problems of aging population, serious imbalance of medical resources supply and demand, as well as uneven geographical distribution, resulting in a huge demand for remote e-health. Particularly, with invasions of COVID-19, the health of people and even social stability have been challenged unprecedentedly. To contribute to these urgent problems, this article proposes a general architecture of the remote e-health, where the city hospital provides the technical supports and services for remote hospitals. Meanwhile, 5G technologies supported telemedicine is introduced to satisfy the high-speed transmission of massive multimedia medical data, and further realize the sharing of medical resources. Moreover, to turn passivity into initiative to prevent COVID-19, a broad area epidemic prevention and control scheme is also investigated, especially for the remote areas. We discuss their principles and key features, and foresee the challenges, opportunities, and future research trends. Finally, a node value and content popularity based caching strategy is introduced to provide a preliminary solution of the massive data storage and low-latency transmission.

cs.NI↗

arXiv2020

Waveguide integrated high performance magneto-optical isolators and circulators on silicon nitride platforms

Optical isolators and circulators are indispensable for photonic integrated circuits (PICs). Despite of significant progress in silicon-on-insulator (SOI) platforms, integrated optical isolators and circulators have been rarely reported on silicon nitride (SiN) platforms. In this paper, we report monolithic integration of magneto-optical (MO) isolators on SiN platforms with record high performances based on standard silicon photonics foundry process and magneto-optical thin film deposition. We successfully grow high quality MO garnet thin films on SiN with large Faraday rotation up to -5900 deg/cm. We show a superior magneto-optical figure of merit (FoM) of MO/SiN waveguides compared to that of MO/SOI in an optimized device design. We demonstrate TM/TE mode broadband and narrow band optical isolators and circulators on SiN with high isolation ratio, low cross talk and low insertion loss. In particular, we observe 1 dB insertion loss and 28 dB isolation ratio in a SiN racetrack resonator-based isolator at 1570.2 nm wavelength. The low thermo-optic coefficient of SiN also ensures excellent temperature stability of the device. Our work paves the way for integration of high performance nonreciprocal photonic devices on SiN platforms.

physics.app-ph↗

arXiv2020

NASE: Learning Knowledge Graph Embedding for Link Prediction via Neural Architecture Search

Link prediction is the task of predicting missing connections between entities in the knowledge graph (KG). While various forms of models are proposed for the link prediction task, most of them are designed based on a few known relation patterns in several well-known datasets. Due to the diversity and complexity nature of the real-world KGs, it is inherently difficult to design a model that fits all datasets well. To address this issue, previous work has tried to use Automated Machine Learning (AutoML) to search for the best model for a given dataset. However, their search space is limited only to bilinear model families. In this paper, we propose a novel Neural Architecture Search (NAS) framework for the link prediction task. First, the embeddings of the input triplet are refined by the Representation Search Module. Then, the prediction score is searched within the Score Function Search Module. This framework entails a more general search space, which enables us to take advantage of several mainstream model families, and thus it can potentially achieve better performance. We relax the search space to be continuous so that the architecture can be optimized efficiently using gradient-based search strategies. Experimental results on several benchmark datasets demonstrate the effectiveness of our method compared with several state-of-the-art approaches.

cs.CL↗

arXiv2020

Dense Small Satellite Networks for Modern Terrestrial Communication Systems: Benefits, Infrastructure, and Technologies

Dense small satellite networks (DSSN) in low earth orbits (LEO) can benefit several mobile terrestrial communication systems (MTCS). However, the potential benefits can only be achieved through careful consideration of DSSN infrastructure and identification of suitable DSSN technologies. In this paper, we discuss several components of DSSN infrastructure including satellite formations, orbital paths, inter-satellite communication (ISC) links, and communication architectures for data delivery from source to destination. We also review important technologies for DSSN as well as the challenges involved in the use of these technologies in DSSN. Several open research directions to enhance the benefits of DSSN for MTCS are also identified in the paper. A case study showing the integration benefits of DSSN in MTCS is also included.

eess.SP↗

arXiv2020

Perpetual Motion: Generating Unbounded Human Motion

The modeling of human motion using machine learning methods has been widely studied. In essence it is a time-series modeling problem involving predicting how a person will move in the future given how they moved in the past. Existing methods, however, typically have a short time horizon, predicting a only few frames to a few seconds of human motion. Here we focus on long-term prediction; that is, generating long sequences (potentially infinite) of human motion that is plausible. Furthermore, we do not rely on a long sequence of input motion for conditioning, but rather, can predict how someone will move from as little as a single pose. Such a model has many uses in graphics (video games and crowd animation) and vision (as a prior for human motion estimation or for dataset creation). To address this problem, we propose a model to generate non-deterministic, \textit{ever-changing}, perpetual human motion, in which the global trajectory and the body pose are cross-conditioned. We introduce a novel KL-divergence term with an implicit, unknown, prior. We train this using a heavy-tailed function of the KL divergence of a white-noise Gaussian process, allowing latent sequence temporal dependency. We perform systematic experiments to verify its effectiveness and find that it is superior to baseline methods.

cs.CV↗

arXiv2020

Learning Posterior and Prior for Uncertainty Modeling in Person Re-Identification

Data uncertainty in practical person reID is ubiquitous, hence it requires not only learning the discriminative features, but also modeling the uncertainty based on the input. This paper proposes to learn the sample posterior and the class prior distribution in the latent space, so that not only representative features but also the uncertainty can be built by the model. The prior reflects the distribution of all data in the same class, and it is the trainable model parameters. While the posterior is the probability density of a single sample, so it is actually the feature defined on the input. We assume that both of them are in Gaussian form. To simultaneously model them, we put forward a distribution loss, which measures the KL divergence from the posterior to the priors in the manner of supervised learning. In addition, we assume that the posterior variance, which is essentially the uncertainty, is supposed to have the second-order characteristic. Therefore, a $Σ-$net is proposed to compute it by the high order representation from its input. Extensive experiments have been carried out on Market1501, DukeMTMC, MARS and noisy dataset as well.

cs.CV↗

arXiv2020

Decentralized Blockchain for Privacy-Preserving Large-Scale Contact Tracing

Activity-tracking applications and location-based services using short-range communication (SRC) techniques have been abruptly demanded in the COVID-19 pandemic, especially for automated contact tracing. The attention from both public and policy keeps raising on related practical problems, including \textit{1) how to protect data security and location privacy? 2) how to efficiently and dynamically deploy SRC Internet of Thing (IoT) witnesses to monitor large areas?} To answer these questions, in this paper, we propose a decentralized and permissionless blockchain protocol, named \textit{Bychain}. Specifically, 1) a privacy-preserving SRC protocol for activity-tracking and corresponding generalized block structure is developed, by connecting an interactive zero-knowledge proof protocol and the key escrow mechanism. As a result, connections between personal identity and the ownership of on-chain location information are decoupled. Meanwhile, the owner of the on-chain location data can still claim its ownership without revealing the private key to anyone else. 2) An artificial potential field-based incentive allocation mechanism is proposed to incentivize IoT witnesses to pursue the maximum monitoring coverage deployment. We implemented and evaluated the proposed blockchain protocol in the real-world using the Bluetooth 5.0. The storage, CPU utilization, power consumption, time delay, and security of each procedure and performance of activities are analyzed. The experiment and security analysis is shown to provide a real-world performance evaluation.

cs.CR↗

arXiv2020

Intelligent Charging Management of Electric Vehicles Considering Dynamic User Behavior and Renewable Energy: A Stochastic Game Approach

Uncoordinated charging of a rapidly growing number of electric vehicles (EVs) and the uncertainty associated with renewable energy resources may constitute a critical issue for the electric mobility (E-Mobility) in the transportation system especially during peak hours. To overcome this dire scenario, we introduce a stochastic game to study the complex interactions between the power grid and charging stations. In this context, existing studies have not taken into account the dynamics of customers' preference on charging parameters. In reality, however, the choice of the charging parameters may vary over time, as the customers may change their charging preferences. We model this behavior of customers with another stochastic game. Moreover, we define a quality of service (QoS) index to reflect how the charging process influences customers' choices on charging parameters. We also develop an online algorithm to reach the Nash equilibria for both stochastic games. Then, we utilize real data from the California Independent System Operator (CAISO) to evaluate the performance of our proposed algorithm. The results reveal that the electricity cost with the proposed method can result in a saving of about 20% compared to the benchmark method, while also yielding a higher QoS in terms of charging and waiting time. Our results can be employed as guidelines for charging service providers to make efficient decisions under uncertainty relative to power generation of renewable energy.

math.OC↗

arXiv2020

Distributed Deep Reinforcement Learning for Intelligent Load Scheduling in Residential Smart Grids

The power consumption of households has been constantly growing over the years. To cope with this growth, intelligent management of the consumption profile of the households is necessary, such that the households can save the electricity bills, and the stress to the power grid during peak hours can be reduced. However, implementing such a method is challenging due to the existence of randomness in the electricity price and the consumption of the appliances. To address this challenge, we employ a model-free method for the households which works with limited information about the uncertain factors. More specifically, the interactions between households and the power grid can be modeled as a non-cooperative stochastic game, where the electricity price is viewed as a stochastic variable. To search for the Nash equilibrium (NE) of the game, we adopt a method based on distributed deep reinforcement learning. Also, the proposed method can preserve the privacy of the households. We then utilize real-world data from Pecan Street Inc., which contains the power consumption profile of more than 1; 000 households, to evaluate the performance of the proposed method. In average, the results reveal that we can achieve around 12% reduction on peak-to-average ratio (PAR) and 11% reduction on load variance. With this approach, the operation cost of the power grid and the electricity cost of the households can be reduced.

math.OC↗

arXiv2020

Intervention scenarios to enhance knowledge transfer in a network of firm

We investigate a multi-agent model of firms in an R\&D network. Each firm is characterized by its knowledge stock $x_{i}(t)$, which follows a non-linear dynamics. It can grow with the input from other firms, i.e., by knowledge transfer, and decays otherwise. Maintaining interactions is costly. Firms can leave the network if their expected knowledge growth is not realized, which may cause other firms to also leave the network. The paper discusses two bottom-up intervention scenarios to prevent, reduce, or delay cascades of firms leaving. The first one is based on the formalism of network controllability, in which driver nodes are identified and subsequently incentivized, by reducing their costs. The second one combines node interventions and network interventions. It proposes the controlled removal of a single firm and the random replacement of firms leaving. This allows to generate small cascades, which prevents the occurrence of large cascades. We find that both approaches successfully mitigate cascades and thus improve the resilience of the R\&D network.

physics.soc-ph↗

arXiv2020

Better Set Representations For Relational Reasoning

Incorporating relational reasoning into neural networks has greatly expanded their capabilities and scope. One defining trait of relational reasoning is that it operates on a set of entities, as opposed to standard vector representations. Existing end-to-end approaches typically extract entities from inputs by directly interpreting the latent feature representations as a set. We show that these approaches do not respect set permutational invariance and thus have fundamental representational limitations. To resolve this limitation, we propose a simple and general network module called a Set Refiner Network (SRN). We first use synthetic image experiments to demonstrate how our approach effectively decomposes objects without explicit supervision. Then, we insert our module into existing relational reasoning models and show that respecting set invariance leads to substantial gains in prediction performance and robustness on several relational reasoning tasks.

cs.LG↗

arXiv2020

Placement and Routing Optimization for Automated Inspection with UAVs: A Study in Offshore Wind Farm

Wind power is a clean and widely deployed alternative to reducing our dependence on fossil fuel power generation. Under this trend, more turbines will be installed in wind farms. However, the inspection of the turbines in an offshore wind farm is a challenging task because of the harsh environment (e.g., rough sea, strong wind, and so on) that leads to high risk for workers who need to work at considerable height. Also, inspecting increasing number of turbines requires long man hours. In this regard, unmanned aerial vehicles (UAVs) can play an important role for automated inspection of the turbines for the operator, thus reducing the inspection time, man hours, and correspondingly the risk for the workers. In this case, the optimal number of UAVs enough to inspect all turbines in the wind farm is a crucial parameter. In addition, finding the optimal path for the UAVs' routes for inspection is also important and is equally challenging. In this paper, we formulate a placement optimization problem to minimize the number of UAVs in the wind farm and a routing optimization problem to minimize the inspection time. Wind has an impact on the flying range and the flying speed of UAVs, which is taken into account for both problems. The formulated problems are NP-hard. We therefore design heuristic algorithms to find solutions to both problems, and then analyze the complexity of the proposed algorithms. The data of the Walney wind farm are then utilized to evaluate the performance of the proposed algorithms. Simulation results clearly show that the proposed methods can obtain the optimal routing path for UAVs during the inspection.

math.OC↗

arXiv2020

Binary linear codes with few weights from Boolean functions

Boolean functions have very nice applications in cryptography and coding theory, which have led to a lot of research focusing on their applications. The objective of this paper is to construct binary linear codes with few weights from the defining set, which is defined by some special Boolean functions and some additional restrictions. First, we provide two general constructions of binary linear codes with three or four weights from Boolean functions with at most three Walsh transform values and determine the parameters of their dual codes. Then many classes of binary linear codes with explicit weight enumerators are obtained. Some binary linear codes and their duals obtained are optimal or almost optimal. The binary linear codes obtained in this paper may have a special interest in secret sharing schemes, association schemes, strongly regular graphs.

cs.IT↗

arXiv2020

Electronic Origin for the Enhanced Thermoelectric Efficiency of Cu2Se

Thermoelectric materials (TMs) can uniquely convert waste heat into electricity, which provides a potential solution for the global energy crisis that is increasingly severe. Bulk Cu2Se, with ionic conductivity of Cu ions, exhibits a significant enhancement of its thermoelectric figure of merit zT by a factor of ~3 near its structural transition around 400 K. Here, we show a systematic study of the electronic structure of Cu2Se and its temperature evolution using high-resolution angle-resolved photoemission spectroscopy. Upon heating across the structural transition, the electronic states near the corner of the Brillouin zone gradually disappear, while the bands near the centre of Brillouin zone shift abruptly towards high binding energies and develop an energy gap. Interestingly, the observed band reconstruction well reproduces the temperature evolution of the Seebeck coefficient of Cu2Se, providing an electronic origin for the drastic enhancement of the thermoelectric performance near 400 K. The current results not only bridge among structural phase transition, electronic structures, and thermoelectric properties in a condensed matter system, but also provide valuable insights into the search and design of new generation of thermoelectric materials.

cond-mat.mtrl-sci↗