Abstract
The expectation that quantum computation might bring performance advantages in machine learning algorithms motivates the work on the quantum versions of artificial neural networks. In this study, we analyse the learning dynamics of a quantum classifier model that works as an open quantum system which is an alternative to the standard quantum circuit model. According to the obtained results, the model can be successfully trained with a gradient descent (GD)-based algorithm. The fact that these optimisation processes have been obtained with continuous dynamics, shows promise for the development of a differentiable activation function for the classifier model.
Similar content being viewed by others
Change history
08 April 2024
A Correction to this paper has been published: https://doi.org/10.1007/s12043-024-02754-x
References
W S McCulloch and W Pitts, Bull. Math. Biophys. 5, 115 (1943)
F Rosenblatt, Psychol. Rev. 65, 386 (1958)
J Misra and I Saha, Neurocomputing 74, 239 (2010)
J Gu, Z Wang, J Kuen, L Ma, A Shahroudy, B Shuai, T Liu, X Wang, G Wang, J Cai and T Chen, Pattern Recognit. 77, 354 (2018)
J Schmidhuber, Neural Netw. 61, 85 (2015)
C H Bennett and D P DiVincenzo, Nature 404, 247 (2000)
A Montanaro, NPJ Quantum Inf. 2, 15023 (2016)
L Banchi, N Pancotti and S Bose, NPJ Quantum Inf. 2, 16019 (2016)
A Y Yamamoto, K M Sundqvist, P Li and H R Harris, Quantum Inf. Process. 17, 128 (2018)
F Tacchino, C Macchiavello, D Gerace and D Bajoni, NPJ Quantum Inf. 5, 1 (2019)
E Torrontegui and J J García-Ripoll, EPL 125, 30004 (2019)
S Mangini, F Tacchino, D Gerace, D Bajoni and C Macchiavello, EPL 134, 10002 (2021)
S Yan, H Qi and W Cui, Phys. Rev. A 102, 052421 (2020)
M Pechal, F Roy, S A Wilkinson, G Salis, M Werninghaus, M J Hartmann and S Filipp, Phys. Rev. Res. 4, 033190 (2022)
T Nguyen, I Paik, Y Watanobe and T C Thang, Electronics 11, 437 (2022)
D Türkpençe, T Ç Akıncı and S Şeker, Phys. Lett. A 383, 1410 (2019)
U Korkmaz and D Türkpençe, Phys. Lett. A 426, 127887 (2022)
U Korkmaz, C Sanga and D Türkpençe, 2021 5th International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT) 105 (2021)
U Korkmaz, C Sanga and D Türkpençe, LNICST 436, 159 (2022)
F Verstraete, M M Wolf and J Ignacio Cirac, Nat. Phys. 5, 633 (2009)
S Deffner and C Jarzynski, Phys. Rev. X 3, 041003 (2013)
S Deffner, Phys. Rev. E 88, 062128 (2013)
M Ziman, P Štelmachovič, V Bužek, M Hillery, V Scarani and N Gisin, Phys. Rev. A 65, 042105 (2002)
R Blume-Kohout and W H Zurek, Found. Phys. 35, 1857 (2005)
M Zwolak and W H Zurek, Phys. Rev. A 95, 030101 (2017)
V Scarani, M Ziman, P Žtelmachovič, N Gisin and V Bužek, Phys. Rev. Lett. 88, 097905 (2002)
D Nagaj, P Štelmachovič, V Bužek and M Kim, Phys. Rev. A 66, 062307 (2002)
P Krantz, M Kjaergaard, F Yan, T P Orlando, S Gustavsson and W D Oliver, Appl. Phys. Rev. 6, 021318 (2019)
K H Wan, O Dahlsten, H Kristjánsson, R Gardner and M S Kim, NPJ Quantum Inf. 3, 36 (2017)
S Ruder, arXiv:1609.04747 [cs] (2017)
M O Scully, M S Zubairy, G S Agarwal and H Walther, Science 299, 862 (2003)
D Karevski and T Platini, Phys. Rev. Lett. 102, 207207 (2009)
Acknowledgements
The authors acknowledge the support from the Scientific and Technological Research Council of Turkey (TÜBİTAK-Grant No. 120F353). The authors also wish to extend special thanks to the Cognitive Systems Lab in the Department of Electrical Engineering for providing the atmosphere for motivational and stimulating discussions.
Author information
Authors and Affiliations
Corresponding author
Additional information
The original online version of this article was revised to retecify the Junk character error in the author name Deniz Türkpençe.
Appendix A. Derivation of the cost function
Appendix A. Derivation of the cost function
In this section, we present the mathematical justifications for numerical calculations in the text. First, we substitute \(\nu =g\) in eq. (19)
and obtain the expression for cost function by taking the partial derivative with respect to the coupling constant g.
In our current example, we have two information reservoirs corresponding to specific magnetisations. Therefore, the actual steady-state magnetisation (eq. (13)) reads as
According to the recipe to derive the cost function, the partial derivatives with respect to \(g_1\) and \(g_2\) separately are obtained as
In our example, the desired magnetisation \(\langle \sigma _z^0\rangle _{\textrm{des}}^{ss}=0.4\) is a constant value in the cost function. Using the expression (A.1), (A.2), (A.3) and (A.4), eq. (18) in expressed as follows:
Next, we substitute \(\nu =\theta \) in eq. (19) as
Regarding eq. (15), one can easily see that the magnetisation of the ith reservoir is \(\langle \sigma _z\rangle _i=\langle \sigma _{i}^+\sigma _{i}^-\rangle - \langle \sigma _{i}^-\sigma _{i}^+\rangle \). Therefore, azimuth parameter-dependent expression of the magnetisation can be easily written as \(\langle \sigma _z\rangle _i=\cos \theta _{i}\).
Equation (A.7) is obtained when we take the partial derivative of the cost function with respect to \(\theta \).
In our current example, we have two information reservoirs corresponding to specific magnetisations. Therefore, the actual steady-state magnetisation (eq. (13)) reads as
According to the recipe to derive the cost function, the partial derivatives with respect to \(\theta _{1}\) and \(\theta _{2}\) separately are obtained as
In our example, the desired magnetisation \(\langle \sigma _z^0\rangle _{\textrm{des}}^{ss}=0\) a constant value in the cost function. Using the expression (A.6), (A.7), (A.8) and (A.9), eq. (A.7) is expressed as follows,
Let us edit eq. (18) for \(\nu =\phi \)
Equation (A.12) is obtained when we take the partial derivative of the cost function with respect to \(\phi \).
In our current example, we have two information reservoirs corresponding to specific magnetisations. Therefore, the actual steady-state magnetisation (eq. (13)) by using eq. (15) reads as
According to the recipe to derive the cost function, the partial derivatives with respect to \(\phi _{1}\) and \(\phi _{2}\) separately are obtained as
In our example, the desired magnetisation is \(\langle \sigma _y^0\rangle _{\textrm{des}}^{ss}=0\) a constant value in the cost function. Using the expression (A.6), (A.12), (A.13) and (A.14)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Korkmaz, U., TÜRKPENÇE, D. Dissipative learning of a quantum classifier. Pramana - J Phys 97, 165 (2023). https://doi.org/10.1007/s12043-023-02653-7
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12043-023-02653-7