Qiang Wang

Qiang Wang

Auto-generated profile, based on publicly available information

Qiang Wang

College of Computer Science and Technology, Zhejiang University, Hangzhou, China

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

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

Representative Publications

View All Publications
8 May 2023 in IEEE Access
Verified
Journal Article
Open access
JLJingru Li
Jingru Li
XCXiaofeng Chen
Xiaofeng Chen
PZPeiyu Zheng
Peiyu Zheng
QWQiang Wang
Qiang Wang
ZYZhi Yu
Zhi Yu
Knowledge Distillation (KD) is designed to train smaller student models using a larger pretrained teacher model. However, in decentralized data systems such as blockchain, privacy concerns may arise, making the data inaccessible. To address this issue, Data-Free KD (DFKD) methods have been proposed, which extract prior knowledge from teacher networks and use it to synthesize data for KD. Previous DFKD methods faced challenges due to the large search space of data generation. Recently, deep generative models (DGMs) have been proposed to learn data distribution using deep networks, which provides an efficient way to reduce the search space by generating a set of pseudo data. In this paper, we explore the performance of KD trained using pseudo samples generated by pretrained DGMs and find that the correlation with image quality is not always positive. Based on this observation, we propose a new DFKD framework called Generative Knowledge Distillation (GenKD) that reduces the search space by constructing a prior distribution modeled by DGMs for their power of likelihood estimation. Specifically, we use energy-based models (EBM) to generate data from the Maximum Likelihood Estimation (MLE) of the EBM and gradients from downstream KD tasks by policy gradient. We then train the student model using the pretrained teacher model and pseudo samples. We also implement our GenKD framework on several widely-used benchmarks, including CIFAR100, CIFAR10, and SVHN. Our experiments demonstrate that we can generate high-quality pseudo samples quantitatively and qualitatively using GenKD. Additionally, the top-1 accuracy of the student network can approach state-of-the-art (SOTA) DFKD methods trained using fewer pseudo samples and less generation time.

20 January 2022 in IEEE Access
Verified
Journal Article
Open access
CGChenglong Gao
Chenglong Gao
KCKai Chen
Kai Chen
QWQiang Wang
Qiang Wang
ZCZhixian Chen
Zhixian Chen
Public-key dual-receiver encryption (PK-DRE) is a kind of particular public-key encryption for enabling two independent recipients to obtain the same plaintext from the same ciphertext. Due to its dual-receiver property, PK-DRE is quite helpful in many scenarios, such as deniable authentication, global key escrow, security puzzle, and even blockchain. To the best of our knowledge, the PK-DRE scheme proposed by Chow, Franklin, and Zhang at CT-RSA 2014 is the best one among the existing schemes in terms of security, public verifiability, and key size. In this paper, we restudy their scheme and find a new security proof that leads to a variant. The resulting PK-DRE scheme is more efficient than the original scheme in terms of ciphertext size and encryption/decryption cost, without losing the CCA security in the standard model and public verifiability.

1 June 2021 in IOP Conference Series: Earth and Environmental Science
Verified
Conference Paper
Open access
PDF file
WQWang Qiang
Wang Qiang
LHLiu Haiying
Liu Haiying
WPWu Ping
Wu Ping
GHGao Haiwen
Gao Haiwen
WJWang Jiaxin
Wang Jiaxin
LXLiu Xin
Liu Xin
Vertical integration is one of the important means to achieve coal-electricity joint ventures. According to the comprehensive analysis of domestic and foreign literature, based on the field survey, nine key drivers of vertical integration of coal-electric energy supply chain are proposed. This basis, using interpretative structural modelling (ISM), a five-level hierarchical structure model of the vertical integration motivation of the coal-electric energy supply chain is constructed, which clearly characterization of the hierarchical relationship between the various drivers. Using MICMAC, find the independent and dependent factors. Furthermore, it reveals the most direct motivation and the most fundamental motivation for driving the vertical integration of coal-electric energy supply chain. At the same time, it can clearly show the role of each driver in the vertical integration of coal-fired energy supply chain.

17 March 2023
Verified
Conference Paper
Open access
AWAn Wen
An Wen
YYYina Yang
Yina Yang
RXRuqiao Xu
Ruqiao Xu
JZJingyi Zhu
Jingyi Zhu
QWQiang Wang
Qiang Wang
Deep neural network is subject to catastrophic forgetting when learning new data incrementally. In this work we propose a class incremental learning algorithm based on feature replay and multi-layer spatial distillation to address the problem of data privacy and catastrophic forgetting. Firstly, different from general methods which save and replay images to retain old knowledge, this method saves prototype and generates feature for replay based on a new feature distribution assumption. Secondly, this method makes better use of the middle layer feature maps and performs multilayer spatial distillation, which alleviates catastrophic forgetting and improves the overall effect of class incremental learning. We implement multiple sets of experiments on public datasets and the experimental results show that the algorithm proposed in this paper can effectively improve the effect of class-incremental learning.


16 June 2023
Verified
Conference Paper
Closed access
PDF file
QZQian Zheng
Qian Zheng
YYYibo Yong
Yibo Yong
QWQiang Wang
Qiang Wang
TZTianqi Zhong
Tianqi Zhong
Tao ZhangTao Zhang
Tao Zhang
QYQiang Yang
Qiang Yang
The standard clauses used in the electric power industry often contain inconsistencies, making it difficult for professionals to select appropriate terms in their work. In this paper, we propose a discrepancy discrimination model based on knowledge graph and natural language processing, using standard clauses of power equipment as the dataset. Furthermore, we construct an ensemble learning model to assist decision-making in the electric power field. The experimental results show that the proposed method is effective in terms of precision and recall. The discrepancy discrimination model has a precision rate of 76.45% and recall rate of 84.72%, while the auxiliary decision-making model has a precision rate of 75.3%. The method is able to effectively identify inconsistencies in equipment standards and provide decision-making advice, providing a valuable reference for standardizing the content of standard terms in the industry.