Yi Ding

Dr. Yi Ding

Beijing Wuzi University

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

2 March 2021 in IEEE Access
Verified
Journal Article
Open access
WXWenyuan Xu
Wenyuan Xu
WFWeiwei Fang
Weiwei Fang
YDYi Ding
Yi Ding
MZMeixia Zou
Meixia Zou
Neal XiongNeal Xiong
Neal Xiong
The ever-increasing number of Internet of Things (IoT) devices are continuously generating huge masses of data, but the current cloud-centric approach for IoT big data analysis has raised public concerns on both data privacy and network cost. Federated learning (FL) recently emerges as a promising technique to accommodate these concerns, by means of learning a global model by aggregating local updates from multiple devices without sharing the privacy-sensitive data. However, IoT devices usually have constrained computation resources and poor network connections, making it infeasible or very slow to train deep neural networks (DNNs) by following the FL pattern. To address this problem, we propose a new efficient FL framework called FL-PQSU in this paper. It is composed of 3-stage pipeline: structured pruning, weight quantization and selective updating, that work together to reduce the costs of computation, storage, and communication to accelerate the FL training process. We study FL-PQSU using popular DNN models (AlexNet, VGG16) and publicly available datasets (MNIST, CIFAR10), and demonstrate that it can well control the training overhead while still guaranteeing the learning performance.

2 December 2020 in Electronics
Verified
Journal Article
Open access
PDF file
KHKai Hu
Kai Hu
Jian ZhuJian Zhu
Jian Zhu
YDYi Ding
Yi Ding
XBXiaomin Bai
Xiaomin Bai
JHjiehua huang
jiehua huang
A smart contract is the algorithmic description of a contractual transaction protocol that is automatically executed together with the information provided by its parties. It is written in a simplified programming language that is specific to a particular domain. Not only correctness and unambiguity are its essential formal properties, but also conformance to any legislation governing the matter of the transaction. Finally, and importantly, the trustworthiness, safety and security of the platform executing the transactions are its main attributes. An emerging challenge is to define a proper engineering process to meet the demanding requirements while supporting mass production and distribution. This paper proposes the concept of smart contract engineering (SCE) to facilitate the generation of smart legal contracts, which is the combination of software engineering, formal methods and computational law. SCE aims to reduce the potential errors and improve efficiency during the contract development process, meanwhile promote the standardization of contract design methodologies. In this paper, the roadmap of an iterative refinement-based, model-driven formal design methodology is introduced, not only to validate smart contracts but also to support the whole life cycle of their engineering.

15 December 2020 in IEEE Transactions on Industrial Informatics
Verified
Journal Article
Closed access
WFWeiwei Fang
Weiwei Fang
FXFeng Xue
Feng Xue
YDYi Ding
Yi Ding
Neal XiongNeal Xiong
Neal Xiong
Victor C. M. LeungVictor C. M. Leung
Victor C. M. Leung
Motivated by the prospects of 5G communications and industrial Internet of Things (IoT), recent years have seen the rise of a new computing paradigm, edge computing, which shifts data analytics to network edges that are at the proximity of big data sources. Although deep neural networks (DNNs) have been extensively used in many platforms and scenarios, they are usually both compute and memory intensive, thus, difficult to be deployed on resource-limited edge devices and in performance-demanding edge applications. Hence, there is an urgent need for techniques that enable DNN models to fit into edge devices, while ensuring acceptable execution costs and inference accuracy. This article proposes an on-demand DNN model inference system for industrial edge devices, called knowledge distillation and early exit on edge (EdgeKE). It focuses on the following two design knobs: first, DNN compression based on knowledge distillation, which trains the compact edge models under the supervision of large complex models for improving accuracy and speed; second, DNN acceleration based on early exit, which provides flexible choices for satisfying distinct latency or accuracy requirements from edge applications. By extensive evaluations on the CIFAR100 dataset and across three state-of-art edge devices, experimental results demonstrate that EdgeKE significantly outperforms the baseline models in terms of inference latency and memory footprint, while maintaining competitive classification accuracy. Furthermore, EdgeKE is verified to be efficiently adaptive to the application requirements on the inference performance. The accuracy loss is within 4.84% under various latency constraints, and the speedup ratio is up to 3.30× under various accuracy requirements.

12 August 2019 in IEEE Access
Verified
Journal Article
Open access
Haitao ZhaoHaitao Zhao
Haitao Zhao
YDYi Ding
Yi Ding
ZMZhang Mengkang
Zhang Mengkang
QWQin Wang
Qin Wang
SXShi Xinyue
Shi Xinyue
HZHongbo Zhu
Hongbo Zhu
With the rapid development of intelligent transportation, various applications which have millisecond delay requirements appear in the vehicular heterogeneous network. Offloading these delay-sensitive applications into edge nodes is a trend and direction of development. However, with the increase in the number of vehicular applications, the conventional methods of distance-based edge node workload allocation make the workload allocation unbalanced, causing some edge nodes to be overloaded, resulting in response time of corresponding application being too long to meet low delay requirements. In this paper, we tackle the problem of edge node overload and propose a multipath transmission workload balancing optimization scheme, which uses multipath transmission in the edge computing architecture as the transport protocol support for the communications between the vehicles and the edge nodes and the real-time virtual machine (VM) migration happens between the edge nodes. First, the application is assigned to the edge node closest to each vehicle. When the workload of the edge node exceeds its capacity, the scheme iteratively selects the application with the longest response time. Second. if its response time exceeds the response time passed to the cloud computing center, it is assigned to the cloud computing center for processing; if not, it is reassigned to the standby edge to minimize its response time until all vehicle applications cannot find a better edge node. Then, computing resources are allocated to each edge node, and resources of different sizes are allocated to different types of VMs in each edge node through convex optimization. Finally, the extensive simulation results illustrate that the multipath transmission workload balancing optimization scheme can effectively reduce the average response time of the vehicular applications compared to existing schemes.

21 December 2020 in IEEE Access
Verified
Journal Article
Open access
YDYi Ding
Yi Ding
CWChenshuo Wang
Chenshuo Wang
QZQionghui Zhong
Qionghui Zhong
HLHaisheng LI
Haisheng LI
JTJinjing Tan
Jinjing Tan
JLJie Li
Jie Li
The close integration of blockchain and smart contract technology has become an important foundation for current trusted applications. High-quality, high-efficiency and high-security codes have become basic requirements for smart contract applications because they are not easy to be modified after being deployed on blockchain. This paper proposes a function-level dynamic monitoring and analysis method for smart contract, and implements a prototype system. The method adds a “shadow stack” and related data structures to virtual machine of testing blockchain platform by analyzing the principle of function management with original stack, then monitors the bytecode after code instrumentation, records the function calling relationships as well as the relevant metrics of time, instruction number and gas consumption. The prototype system identifies contract inefficient behaviors using visualization and intelligent analysis methods, then forms a smart contract optimization closed loop through iterative improvement. Finally, the paper verified the high feasibility and applicability of the monitoring and analyzing method as well as prototype system’s performance through experiments.