Zhehao Huang

Xiangtan University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Xiaolin Huang

Shanghai Jiao Tong University

Qinghua Tao

KU Leuven

Sizhe Chen

State Key Laboratory of Remote Sensing Science

Yipeng Liu

University of Electronic Science and Technology of China

Lidian Chen

Hunan Provincial Center for Disease Control and Prevention

Tao Li

Southwest Jiaotong University

Cunyi Yang

South China Agricultural University

Benhuan Nie

Guangzhou University

Tao Li

Hainan University

B Luo

Chongqing University of Science and Technology

Cooperative Institutions

Shanghai Jiao Tong University

Xiangtan University

Lanzhou University

Nanjing University of Aeronautics and Astronautics

Chongqing University of Posts and Telecommunications

University of Chinese Academy of Sciences

Kunming University of Science and Technology

South China Agricultural University

Guangzhou University

University of Electronic Science and Technology of China

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Query Attack by Multi-Identity Surrogates

IEEE Transactions on Artificial Intelligence (2023)

Sizhe Chen Zhehao Huang Qinghua Tao Xiaolin Huang

Deep neural networks (DNNs) are acknowledged as vulnerable to adversarial attacks while the existing black-box attacks require extensive queries on the victim DNN to achieve high success rates. For query-efficiency, surrogate models of the victim are used to generate transferable adversarial examples (AEs) because of their gradient similarity (GS), i.e., surrogates' attack gradients are similar to the victim's ones. However, it is generally neglected to exploit their similarity on outputs, namely the prediction similarity (PS), to filter out inefficient queries by surrogates without querying the victim. To jointly utilize and also optimize surrogates' GS and PS, we develop QueryNet, a unified attack framework that can significantly reduce queries. QueryNet creatively attacks by multi-identity surrogates, i.e., crafts several AEs for one sample by different surrogates and also uses surrogates to decide on the most promising AE for the query. After that, the victim's query feedback is accumulated to optimize not only surrogates' parameters but also their architectures, enhancing both the GS and the PS. Although QueryNet has no access to pretrained surrogates' prior, it reduces queries by averagely about an order of magnitude compared to alternatives within an acceptable time, according to our comprehensive experiments: 11 victims (including two commercial models) on MNIST/CIFAR10/ImageNet, allowing only 8-b image queries, and no access to the victim's training data. The code is available at https://github.com/Sizhe-Chen/QueryNet .

Similarity (geometry)

MNIST database

Pseudonym

Identification

10.1109/tai.2023.3257276

Cite

Citations (1)

QueryNet: An Efficient Attack Framework with Surrogates Carrying Multiple Identities.

arXiv (Cornell University) (2021)

Sizhe Chen Zhehao Huang Qinghua Tao Xiaolin Huang

Deep Neural Networks (DNNs) are acknowledged as vulnerable to adversarial attacks, while the existing black-box attacks require extensive queries on the victim DNN to achieve high success rates. For query-efficiency, surrogate models of the victim are adopted as transferable attackers in consideration of their Gradient Similarity (GS), i.e., surrogates' attack gradients are similar to the victim's ones to some extent. However, it is generally neglected to exploit their similarity on outputs, namely the Prediction Similarity (PS), to filter out inefficient queries. To jointly utilize and also optimize surrogates' GS and PS, we develop QueryNet, an efficient attack network that can significantly reduce queries. QueryNet crafts several transferable Adversarial Examples (AEs) by surrogates, and then decides also by surrogates on the most promising AE, which is then sent to query the victim. That is to say, in QueryNet, surrogates are not only exploited as transferable attackers, but also as transferability evaluators for AEs. The AEs are generated using surrogates' GS and evaluated based on their FS, and therefore, the query results could be back-propagated to optimize surrogates' parameters and also their architectures, enhancing both the GS and the FS. QueryNet has significant query-efficiency, i.e., reduces queries by averagely about an order of magnitude compared to recent SOTA methods according to our comprehensive and real-world experiments: 11 victims (including 2 commercial models) on MNIST/CIFAR10/ImageNet, allowing only 8-bit image queries, and no access to the victim's training data.

MNIST database

Transferability

Similarity (geometry)

Deep Neural Networks

Source

Cite

Citations (0)

Typical Scenario Load Identification Based on Feature Fusion and Transfer Learning

Recent Patents on Mechanical Engineering (2024)

Lingzhi Yi Yuhang Gao Yahui Wang B Luo Zhehao Huang

Background: The electricity demand is continuously increasing. However, various institutions, enterprises, and individuals exhibit many irregularities in their electricity usage, leading to significant wastage of electricity. To achieve effective energy management, researchers are attempting to analyze and regulate users' electricity demands by monitoring their load usage through Non- Intrusive Load Monitoring (NILM) technology. The accuracy of load identification in this technology will greatly impact the results of load monitoring. Although there are currently many articles and patents related to NILM, they utilize a large amount of computational resources and require high sampling rates from devices, yet the results are still unsatisfactory. Therefore, it is necessary to improve the accuracy of load identification in data with relatively low sampling frequencies. Objective: To improve the accuracy of load identification with low sampling frequency data, this paper proposes a typical scenario load identification method based on feature fusion and transfer learning. Methods: This method adopts the fusion of current and power factor angles to provide abundant identification information for NILM, effectively reducing the situation of single-feature overlap of different loads. By inputting the fused feature data into GoogLeNet and utilizing transfer learning for training, not only is the accuracy improved, but also the training time and the requirement for the sampling rate of training data are greatly reduced. In addition, selecting typical scenario loads can monitor loads in a targeted manner, reduce the waste of computing resources caused by irrelevant loads, and more effectively guide electricity usage strategies. Results: The proposed load identification method was tested on the low sampling frequency dataset used in this paper. It achieved an overall load identification accuracy of 94.61% across three scenarios, improving accuracy by 3% to 7% compared to other models. Conclusion: The simulation results indicate that this method achieves high load identification accuracy at low sampling frequencies. It also exhibits good generalization ability. This method not only reduces the performance requirements for monitoring equipment but also enhances monitoring efficiency.

Identification

Feature (linguistics)

Transfer of learning

10.2174/0122127976325514241001045948

Cite

Citations (0)

QueryNet: An Attack Framework with Surrogates Carrying Multiple Identities.

arXiv (Cornell University) (2021)

Sizhe Chen Zhehao Huang Qinghua Tao Xiaolin Huang

Deep Neural Networks (DNNs) are acknowledged as vulnerable to adversarial attacks, while the existing black-box attacks require extensive queries on the victim DNN to achieve high success rates. For query-efficiency, surrogate models of the victim are adopted as transferable attackers in consideration of their Gradient Similarity (GS), i.e., surrogates' attack gradients are similar to the victim's ones to some extent. However, it is generally neglected to exploit their similarity on outputs, namely the Prediction Similarity (PS), to filter out inefficient queries. To jointly utilize and also optimize surrogates' GS and PS, we develop QueryNet, an efficient attack network that can significantly reduce queries. QueryNet crafts several transferable Adversarial Examples (AEs) by surrogates, and then decides also by surrogates on the most promising AE, which is then sent to query the victim. That is to say, in QueryNet, surrogates are not only exploited as transferable attackers, but also as transferability evaluators for AEs. The AEs are generated using surrogates' GS and evaluated based on their PS, and therefore, the query results could be back-propagated to optimize surrogates' parameters and also their architectures, enhancing both the GS and the PS. QueryNet has significant query-efficiency, i.e., reduces queries by averagely about an order of magnitude compared to recent SOTA methods according to our comprehensive and real-world experiments: 11 victims (including 2 commercial models) on MNIST/CIFAR10/ImageNet, allowing only 8-bit image queries, and no access to the victim's training data.

Transferability

MNIST database

Similarity (geometry)

Deep Neural Networks

Source

Cite

Citations (1)

Online Continual Learning via Logit Adjusted Softmax

arXiv (Cornell University) (2023)

Zhehao Huang Tao Li Chenhe Yuan Yingwen Wu Xiaolin Huang

Online continual learning is a challenging problem where models must learn from a non-stationary data stream while avoiding catastrophic forgetting. Inter-class imbalance during training has been identified as a major cause of forgetting, leading to model prediction bias towards recently learned classes. In this paper, we theoretically analyze that inter-class imbalance is entirely attributed to imbalanced class-priors, and the function learned from intra-class intrinsic distributions is the Bayes-optimal classifier. To that end, we present that a simple adjustment of model logits during training can effectively resist prior class bias and pursue the corresponding Bayes-optimum. Our proposed method, Logit Adjusted Softmax, can mitigate the impact of inter-class imbalance not only in class-incremental but also in realistic general setups, with little additional computational cost. We evaluate our approach on various benchmarks and demonstrate significant performance improvements compared to prior arts. For example, our approach improves the best baseline by 4.6% on CIFAR10.

Softmax function

10.48550/arxiv.2311.06460

Cite

Citations (0)

Corporate financialization matters trade credit financing: Evidence from China’s non-financial listed companies

Economic Analysis Letters (2024)

Xiaohua Huang Benhuan Nie Zhehao Huang

This study empirically examines the influence of corporate financialization on trade credit financing, focusing on non-financial companies listed in China's A-share market from 2010 to 2022. The results indicate that as corporate financialization increases, companies will obtain less trade credit financing. Further incorporating moderating factors like market position and financing constraints, we find that a higher market position effectively mitigates the negative impact of corporate financialization on trade credit financing, whereas higher financing constraints exacerbate the adverse effects of corporate financialization on trade credit financing. Additional analysis shows that corporate financialization negatively affects trade credit financing by crowding out main business performance and increasing operational risks. Moreover, corporate financialization more strongly inhibits trade credit financing in state-owned companies.

Financialization

Trade Credit

Corporate Finance

10.58567/eal03030004

Cite

Citations (0)

Low Dimensional Trajectory Hypothesis is True: DNNs can be Trained in Tiny Subspaces

IEEE Transactions on Pattern Analysis and Machine Intelligence (2022)

Tao Li Lei Tan Zhehao Huang Qinghua Tao Yipeng Liu

Deep neural networks (DNNs) usually contain massive parameters, but there is redundancy such that it is guessed that they could be trained in low-dimensional subspaces. In this paper, we propose a Dynamic Linear Dimensionality Reduction (DLDR) based on the low-dimensional properties of the training trajectory. The reduction method is efficient, supported by comprehensive experiments: optimizing DNNs in 40-dimensional spaces can achieve comparable performance as regular training over thousands or even millions of parameters. Since there are only a few variables to optimize, we develop an efficient quasi-Newton-based algorithm, obtain robustness to label noise, and improve the performance of well-trained models, which are three follow-up experiments that can show the advantages of finding such low-dimensional subspaces. The code is released (Pytorch: https://github.com/nblt/DLDR and Mindspore: https://gitee.com/mindspore/docs/tree/r1.6/docs/sample_code/dimension_reduce_training).

Robustness

Deep Neural Networks

Code (set theory)

10.1109/tpami.2022.3178101

Cite

Citations (15)

The Dark Side of Financial Digitalization: Corporate Digital Finance and Speculative Financial Investments

Zhehao Huang Cunyi Yang Lidian Chen

Corporate Finance

Great Rift

10.2139/ssrn.5018202

Cite

Citations (0)