Abstract
It has been well known that there is an intriguing analogy between statistical mechanics of Ising spin glasses and information processing such as the associative memory, the image restoration, the pattern recognition, and etc. As a natural extension of the statistical mechanical model of information processing, one can consider a quantum mechanical model of information processing using transverse Ising models. ChapterĀ 9 discusses such an application of transverse Ising models to information processing. The first part focuses on the Hopfield model in a transverse field for the associative memory, where the extraction of embedded patterns using quantum fluctuations is discussed. The second part of this chapter is devoted to the investigation of image restorations and error correcting codes by Bayesian statistics with quantum fluctuations. It mentions the MPM estimate as well the MAP estimate which respectively correspond to the Bayesian estimation schemes by tuning the amount of quantum fluctuations and by quantum annealing. Appendix includes details of the Hopfield model in a transverse field.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Amara, P., Hsu, D., Straub, J.E.: Global energy minimum searches using an approximate solution of the imaginary time Schroedinger equation. J. Phys. Chem. 97(25), 6715ā6721 (1993). [8.1, 9.2]
Amit, D.J., Gutfreund, H., Sompolinsky, H.: Storing infinite numbers of patterns in a spin-glass model of neural networks. Phys. Rev. Lett. 55, 1530ā1533 (1985). [9.1.1]
Ash, R.B.: Information Theory. Dover, New York (1965). [9.2.2, 9.2.4]
Chakrabarti, B.K., Dasgupta, P.K.: Modelling neural networks. Phys. A, Stat. Mech. Appl. 186, 33ā48 (1992). [9.1.1]
Chakrabarti, B.K., Dutta, A., Sen, P.: Quantum Ising Phases and Transitions in Transverse Ising Models. Springer, Berlin (1995). [9.2]
Coolen, A.C.C., Ruijgrok, T.W.: Image evolution in Hopfield networks. Phys. Rev. A 38, 4253ā4255 (1988). [9.1.2]
Das, A., Chakrabarti, B.K.: Quantum Annealing and Related Optimization Methods. Lecture Notes in Physics, vol. 679. Springer, Berlin (2005). [1.3, 8.1, 9.2]
de Almeida, J.R.L., Thouless, D.J.: Stability of the Sherrington-Kirkpatrick solution of a spin glass model. J. Phys. A, Math. Gen. 11(5), 983 (1978). [9.1.1]
Finnila, A., Gomez, M., Sebenik, C., Stenson, C., Doll, J.: Quantum annealing: a new method for minimizing multidimensional functions. Chem. Phys. Lett. 219(5ā6), 343ā348 (1994). [8.1, 9.2]
Geman, S., Geman, D.: Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. IEEE Trans. Pattern Anal. Mach. Intell. PAMI-6(6), 721ā741 (1984). [8.1, 8.7.2, 8.A.3, 9.2, 9.2.1]
Goldschmidt, Y.Y.: Solvable model of the quantum spin glass in a transverse field. Phys. Rev. B 41, 4858ā4861 (1990). [1.3, 6.6, 8.5.3.2, 9.2]
Hopfield, J.: Neural networks and physical systems with emergent collective computational abilities. Proc. Natl. Acad. Sci. USA 79(8), 2554ā2558 (1982). [9.1]
Inoue, J.: Application of the quantum spin glass theory to image restoration. Phys. Rev. E 63, 046114 (2001). [1.3, 9.2, 9.2.4]
Inoue, J.: Pattern-recalling processes in quantum Hopfield networks far from saturation. J.Ā Phys. Conf. Ser. 297(1), 012012 (2011). [1.1, 1.3, 9.1.2]
Inoue, J., Tanaka, K.: Dynamics of the maximum marginal likelihood hyperparameter estimation in image restoration: gradient descent versus expectation and maximization algorithm. Phys. Rev. E 65, 016125 (2001). [9.2, 9.2.4]
Inoue, J., Saika, Y., Okada, M.: Quantum mean-field decoding algorithm for error-correcting codes. J. Phys. Conf. Ser. 143(1), 012019 (2009). [9.2.5]
Ishii, H., Yamamoto, T.: Effect of a transverse field on the spin glass freezing in the Sherrington-Kirkpatrick model. J. Phys. C, Solid State Phys. 18(33), 6225 (1985). [1.3, 6.2, 6.3, 9.2.5]
Jordan, M.: Learning in Graphical Models. MIT Press, Cambridge (1998). [9.2.5]
Kabashima, Y., Saad, D.: Statistical mechanics of error-correcting codes. Europhys. Lett. 45(1), 97 (1999). [9.2]
Kadowaki, T., Nishimori, H.: Quantum annealing in the transverse Ising model. Phys. Rev. E 58, 5355ā5363 (1998). [1.1, 1.3, 8.1, 8.4.1, 9.2, 9.2.6]
Kirkpatrick, S., Gelatt, C.D., Vecchi, M.P.: Optimization by simulated annealing. Science 220(4598), 671ā680 (1983). [8.1, 9.2, 9.2.1]
Ma, Y.q., Gong, C.d.: Statics in the random quantum asymmetric Sherrington-Kirkpatrick model. Phys. Rev. B 45, 793ā796 (1992). [1.3, 9.1.1]
Ma, Y.q., Gong, C.d.: Hopfield spin-glass model in a transverse field. Phys. Rev. B 48, 12778ā12782 (1993). [1.3,9.1.1]
Ma, Y.q., Zhang, Y.m., Ma, Y.g., Gong, C.d.: Statistical mechanics of a Hopfield neural-network model in a transverse field. Phys. Rev. E 47, 3985ā3987 (1993). [1.3, 9.1.1]
MacKay, D.J.C.: Information Theory, Inference, and Learning Algorithms. Cambridge University Press, Cambridge (2003). [8.1, 9.2.1, 9.2.4]
Miyashita, S.: Dynamics of the magnetization with an inversion of the magnetic field. J.Ā Phys. Soc. Jpn. 64(9), 3207ā3214 (1995). [9.2]
Miyashita, S.: Observation of the energy gap due to the quantum tunneling making use of the Landau-Zener mechanism. J. Phys. Soc. Jpn. 65(8), 2734ā2735 (1996). [9.2]
Morita, S., Nishimori, H.: Mathematical foundation of quantum annealing. J. Math. Phys. 49(12), 125210 (2008). [8.8, 9.2.6]
Nakano, K.: Associatronāa model of associative memory. IEEE Trans. Syst. Man Cybern. 2(3), 380ā388 (1972). [9.1]
Nishimori, H.: Optimum decoding temperature for error-correcting codes. J. Phys. Soc. Jpn. 62(9), 2973ā2975 (1993). [9.2, 9.2.1]
Nishimori, H.: Statistical Physics of Spin Glasses and Information Processing: An Introduction. Oxford University Press, Oxford (2001). [1.3, 8.1, 9.2, 9.2.2]
Nishimori, H., Nonomura, Y.: Quantum effects in neural networks. J. Phys. Soc. Jpn. 65(12), 3780ā3796 (1996). [1.1, 1.3, 9.1.1, 9.A]
Nishimori, H., Wong, K.Y.M.: Statistical mechanics of image restoration and error-correcting codes. Phys. Rev. E 60, 132ā144 (1999). [9.2, 9.2.1, 9.2.4]
Opper, M., Saad, D.: Advanced Mean Field Methods: Theory and Practice. MIT Press, Cambridge (2001). [9.2.5]
Penrose, R.: Shadows of the Mind. Oxford University Press, Oxford (1994). [9.1.1]
Pryce, J.M., Bruce, A.D.: Statistical mechanics of image restoration. J. Phys. A, Math. Gen. 28(3), 511 (1995). [9.2]
RujĆ”n, P.: Finite temperature error-correcting codes. Phys. Rev. Lett. 70, 2968ā2971 (1993). [9.2, 9.2.1]
Santoro, G.E., MartoÅĆ”k, R., Tosatti, E., Car, R.: Theory of quantum annealing of an Ising spin glass. Science 295(5564), 2427ā2430 (2002). [8.4.1, 9.2]
Sherrington, D., Kirkpatrick, S.: Solvable model of a spin-glass. Phys. Rev. Lett. 35, 1792ā1796 (1975). [6.1, 9.1.1, 9.1.2, 9.2, 9.2.4, 9.2.5]
Sourlas, N.: Spin-glass models as error-correcting codes. Nature 339, 693ā695 (1989). [9.2, 9.2.2, 9.2.4]
Suzuki, M.: Relationship between d-dimensional quantal spin systems and (d+1)-dimensional Ising systems. Prog. Theor. Phys. 56(5), 1454ā1469 (1976). [1.1, 1.3, 3.1, 5.2, 8.7.2, 9.1.2, 9.2, 9.2.4, 9.2.5, 9.2.6]
Tanaka, K.: Statistical-mechanical approach to image processing. J. Phys. A, Math. Gen. 35(37), 81 (2002). [9.2]
Tanaka, K., Horiguchi, T.: Quantum statistical-mechanical iterative method in image restoration. Electron. Commun. Jpn. 83(3), 84 (2000). [1.3, 9.2, 9.2.5]
Trotter, H.F.: On the product of semi-groups of operators. Proc. Am. Math. Soc. 10, 545ā551 (1959). [3.1, 9.2.4]
Vitiello, G.: Coherence and dissipative dynamics in the quantum brain model. Neural Netw. World 5, 717 (1995). [9.1.1]
Winkler, G.: Image Analysis, Random Fields, and Markov Chain Monte Carlo Methods: AĀ Mathematical Introduction. Springer, Berlin (2002). [9.2]
Yamamoto, T., Ishii, H.: A perturbation expansion for the Sherrington-Kirkpatrick model with a transverse field. J. Phys. C, Solid State Phys. 20(35), 6053 (1987). [1.3, 6.2, 6.3, 9.2.5]
Zener, C.: Non-adiabatic crossing of energy levels. Proc. R. Soc. Lond. Ser. A 137, 696ā702 (1932). [7.A.2, 9.2]
Author information
Authors and Affiliations
Appendix 9.A: Derivation of Saddle Point Equations forĀ theĀ Quantum Hopfield Model
Appendix 9.A: Derivation of Saddle Point Equations forĀ theĀ Quantum Hopfield Model
In this appendix, according to Nishimori and Nonomura [298], we derive the saddle point equations (9.1.9)ā(9.1.15) for the Hopfield model at zero temperature. We starts our argument from the Suzuki-Trotter decomposition:
where we defined the two parts of the Hamiltonian as
Z M is given by
with
To average out the pattern-dependence in the self-averaging quantity such as free energy, we use the replica trick:
Then, we have
Here we assume that for the first s patterns, namely, for the so-called ācondensed patternsā, the quantity \((\beta/M) \sum_{K\rho} \sum_{\mu=1}^{s} m_{K \mu\rho} \sum_{i}\xi_{i}^{\mu} S_{i\sigma K}\) is of order 1 object, whereas for the un-condensed patterns Ī¼=s,ā¦,p, the quantity \((\beta/M) \sum_{K \rho} \sum_{\mu=s+1}^{P} \sum_{i}\xi_{i}^{\mu} S_{i\rho K}\) is of order \(1/\sqrt{N}\). Hence, we can rewrite (9.A.6) as
Collecting the quadratic part in (9.A.6), we obtain
where we defined
with (MnĆMn)-matrix:
and
Thus, we immediately obtain
where Ī» KĻ are the eigenvalues of the Ī KĻ,LĻ . Inserting the definitions (9.A.11) by using the identities:
We write the factor of unity G{Ī» KĻ }=1 as
where we used the Fourier transform of the delta function. Multiplying the G{Ī» KĻ }=1 by āŖZ nā«, one obtains
where we used self-averaging properties in the limit of Nāā and dropped the site i dependence. f denotes the free energy per spin, which is given by
where T is temperature T=Ī² ā1. Extremum conditions for the free energy give the following saddle points:
where we should notice that m KĪ¼Ļ , q ĻĻ (KL) and r ĻĻ (KL) are overlap, spin glass order parameter and noise from un-condensed patterns, respectively. These are the same as in the classical Hopfield model. On the other hand, \(\overline{S}_{\rho}(KL)\) and t Ļ (KL) represent quantum effects. The former represents the degree of quantum fluctuation by the deviation from unity and the latter denotes the effect of un-correlated patterns in the same replica.
9.1.1 9.A.1 Replica Symmetric and Static Approximation
To proceed the calculations, we assume the replica symmetric and static approximations:
namely, we consider the case in which these order parameters are independent on the replica and Trotter indexes. Then, we have the free energy per replica as follows
By using the Hubbard-Stratonovich transformation, we obtain in the limit of nā0 as
We next consider the term (1/n)logĪ» KĻ in (9.A.28). The matrix elements {Ī KĻ,LĻ } are āĪ²/M (Ļā Ļ), \(-\beta\overline{S}/M\) (Ļ=Ļ, Kā L) and 1āĪ²/M (Ļ=Ļ, K=L). Then, the eigenvalues are given by
Thus, one can evaluate
Collecting therms in (9.A.28), (9.A.29) and (9.A.33), we obtain
For simplicity, we now set m Ī¼ =Ī“ Ī¼,1 m,Ī¾ Ī¼ =Ī“ Ī¼,1, namely, we consider the case in which a single pattern is recalled. The extremum conditions gives the following saddle point equations.
where we defined
9.1.2 9.A.2 Zero Temperature Limit
At the ground state Tā0, we naturally expect that \(\overline{S}=q\). For convenience, we introduce the parameter C as
Then, we can rewrite the saddle point equations in terms of C as
Rights and permissions
Copyright information
Ā© 2013 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Suzuki, S., Inoue, Ji., Chakrabarti, B.K. (2013). Applications. In: Quantum Ising Phases and Transitions in Transverse Ising Models. Lecture Notes in Physics, vol 862. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-33039-1_9
Download citation
DOI: https://doi.org/10.1007/978-3-642-33039-1_9
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-33038-4
Online ISBN: 978-3-642-33039-1
eBook Packages: Physics and AstronomyPhysics and Astronomy (R0)