Quantum adversarial metric learning model based on triplet loss function

EPJ Quantum Technology

Research

Yan-Yan Hou¹  Chong-Qiang Ye²  Jian Li³  Xiu-Bo Chen⁴ 

[1] College of Information Science and Engineering, ZaoZhuang University, 277160, ZaoZhuang, Shandong, China;School of Artificial Intelligence, Beijing University of Posts and Telecommunications, 100876, Beijing, China;School of Artificial Intelligence, Beijing University of Posts and Telecommunications, 100876, Beijing, China;School of Cyberspace Security, Beijing University of Posts Telecommunications, 100876, Beijing, China;School of Cyberspace Security, Beijing University of Posts Telecommunications, 100876, Beijing, China;Information Security Center, State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, 100876, Beijing, China;

关键词: Metric learning;    Hybrid quantum-classical algorithm;    Quantum machine learning;   

DOI  :  10.1140/epjqt/s40507-023-00182-1

 received in 2022-08-19, accepted in 2023-06-14,  发布年份 2023

来源: Springer

【 摘 要 】

Metric learning plays an essential role in image analysis and classification, and it has attracted more and more attention. In this paper, we propose a quantum adversarial metric learning (QAML) model based on the triplet loss function, where samples are embedded into the high-dimensional Hilbert space and the optimal metric is obtained by minimizing the triplet loss function. The QAML model employs entanglement and interference to build superposition states for triplet samples so that only one parameterized quantum circuit is needed to calculate sample distances, which reduces the demand for quantum resources. Considering the QAML model is fragile to adversarial attacks, an adversarial sample generation strategy is designed based on the quantum gradient ascent method, effectively improving the robustness against the functional adversarial attack. Simulation results show that the QAML model can effectively distinguish samples of MNIST and Iris datasets and has higher ϵ-robustness accuracy over the general quantum metric learning. The QAML model is a fundamental research problem of machine learning. As a subroutine of classification and clustering tasks, the QAML model opens an avenue for exploring quantum advantages in machine learning.

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

41116_2023_37_Article_IEq186.gif

Fig. 1

41116_2023_37_Article_IEq184.gif

Fig. 4

41116_2023_37_Article_IEq183.gif

41116_2023_37_Article_IEq182.gif

41116_2023_37_Article_IEq181.gif

41116_2023_37_Article_IEq180.gif

Fig. 2

41116_2023_37_Article_IEq179.gif

41116_2023_37_Article_IEq178.gif

41116_2023_37_Article_IEq175.gif

41116_2023_37_Article_IEq174.gif

41116_2023_37_Article_IEq173.gif

41116_2023_37_Article_IEq172.gif

41116_2023_37_Article_IEq171.gif

Fig. 2

Fig. 4

41116_2023_37_Article_IEq168.gif

41116_2023_37_Article_IEq167.gif

Fig. 6

Fig. 2

Fig. 2

41116_2023_37_Article_IEq163.gif

Fig.1

41116_2023_37_Article_IEq161.gif

41116_2023_37_Article_IEq160.gif

41116_2023_37_Article_IEq159.gif

41116_2023_37_Article_IEq157.gif

41116_2023_37_Article_IEq153.gif

41116_2023_37_Article_IEq152.gif

41116_2023_37_Article_IEq151.gif

41116_2023_37_Article_IEq150.gif

41116_2023_37_Article_IEq149.gif

41116_2023_37_Article_IEq148.gif

41116_2023_37_Article_IEq147.gif

Fig. 1

41116_2023_37_Article_IEq145.gif

41116_2023_37_Article_IEq144.gif

41116_2023_37_Article_IEq143.gif

41116_2023_37_Article_IEq142.gif

Fig. 2

41116_2023_37_Article_IEq141.gif

41116_2023_37_Article_IEq140.gif

41116_2023_37_Article_IEq139.gif

41116_2023_37_Article_IEq138.gif

41116_2023_37_Article_IEq137.gif

41116_2023_37_Article_IEq136.gif

Fig. 2

41116_2023_37_Article_IEq134.gif

41116_2023_37_Article_IEq133.gif

41116_2023_37_Article_IEq132.gif

Fig. 1

41116_2023_37_Article_IEq130.gif

41116_2023_37_Article_IEq129.gif

41116_2023_37_Article_IEq128.gif

Fig. 1

41116_2023_37_Article_IEq126.gif

41116_2023_37_Article_IEq125.gif

41116_2023_37_Article_IEq124.gif

41116_2023_37_Article_IEq123.gif

41116_2023_37_Article_IEq122.gif

Fig. 1

41116_2023_37_Article_IEq121.gif

41116_2023_37_Article_IEq120.gif

41116_2023_37_Article_IEq119.gif

41116_2023_37_Article_IEq118.gif

42004_2023_919_Article_IEq16.gif

41116_2023_37_Article_IEq117.gif

41116_2023_37_Article_IEq116.gif

41116_2023_37_Article_IEq115.gif

41116_2023_37_Article_IEq112.gif

Fig. 3

41116_2023_37_Article_IEq109.gif

41116_2023_37_Article_IEq108.gif

Fig. 1

41116_2023_37_Article_IEq106.gif

Fig. 3

41116_2023_37_Article_IEq104.gif

41116_2023_37_Article_IEq103.gif

41116_2023_37_Article_IEq102.gif

41116_2023_37_Article_IEq101.gif

41116_2023_37_Article_IEq100.gif

41116_2023_37_Article_IEq99.gif

41116_2023_37_Article_IEq89.gif

41116_2023_37_Article_IEq88.gif

Fig. 1

41116_2023_37_Article_IEq86.gif

41116_2023_37_Article_IEq85.gif

41116_2023_37_Article_IEq83.gif

41116_2023_37_Article_IEq82.gif

Fig. 4

41116_2023_37_Article_IEq80.gif

41116_2023_37_Article_IEq79.gif

41116_2023_37_Article_IEq78.gif

Fig. 3

41116_2023_37_Article_IEq76.gif

41116_2023_37_Article_IEq75.gif

41116_2023_37_Article_IEq74.gif

41116_2023_37_Article_IEq73.gif

41116_2023_37_Article_IEq72.gif

Fig. 2

41116_2023_37_Article_IEq69.gif

Fig. 2

41116_2023_37_Article_IEq67.gif

Fig. 2

41116_2023_37_Article_IEq65.gif

41116_2023_37_Article_IEq64.gif

41116_2023_37_Article_IEq63.gif

41116_2023_37_Article_IEq62.gif

41116_2023_37_Article_IEq61.gif

41116_2023_37_Article_IEq60.gif

41116_2023_37_Article_IEq59.gif

41116_2023_37_Article_IEq58.gif

41116_2023_37_Article_IEq56.gif

41116_2023_37_Article_IEq55.gif

41116_2023_37_Article_IEq54.gif

41116_2023_37_Article_IEq53.gif

41116_2023_37_Article_IEq52.gif

Fig. 1

41116_2023_37_Article_IEq50.gif

41116_2023_37_Article_IEq49.gif

41116_2023_37_Article_IEq39.gif

Fig. 2

41116_2023_37_Article_IEq38.gif

41116_2023_37_Article_IEq37.gif

41116_2023_37_Article_IEq36.gif

Fig. 1

41116_2023_37_Article_IEq34.gif

41116_2023_37_Article_IEq33.gif

41116_2023_37_Article_IEq32.gif

41116_2023_37_Article_IEq30.gif

41116_2023_37_Article_IEq29.gif

41116_2023_37_Article_IEq28.gif

41116_2023_37_Article_IEq27.gif

41116_2023_37_Article_IEq26.gif

Fig. 3

Fig. 2

Fig. 2

41116_2023_37_Article_IEq10.gif

41116_2023_37_Article_IEq9.gif

41116_2023_37_Article_IEq7.gif

41116_2023_37_Article_IEq6.gif

41116_2023_37_Article_IEq4.gif

41116_2023_37_Article_IEq3.gif

Fig. 2

Fig. 4

Fig. 3

Fig. 1

Fig. 2

Fig. 6

Fig. 12

Fig. 1

Fig. 4

Fig. 5

Fig. 6

Fig. 5

Scheme 1

Fig. 1

Fig. 1

Fig. 2

Fig. 1

Fig. 11

Fig. 17

Fig. 10

Fig. 3

Fig. 3

Fig. 3