Some implications of renormalization group theoretical ideas to statistics

Physica D: Nonlinear Phenomena

Volumn 205, Issue 1-4, 2005, Pages 207-214

  Rajaram, S ^a   Taguchi, Y H ^b   Oono, Y ^a,c  

^a University of Illinois at Urbana Champaign   (United States)

^b CHUO UNIVERSITY   (Japan)

^c UNIVERSITY OF ILLINOIS AT URBANA CHAMPAIGN   (United States)

Author keywords

Data analysis; Renormalization; Statistics

Indexed keywords

ALGORITHMS; DATA MINING; DATA REDUCTION; DIFFERENTIAL EQUATIONS; LEARNING SYSTEMS; PATTERN RECOGNITION; PERTURBATION TECHNIQUES; POPULATION STATISTICS; STATISTICS;

ASYMPTOTIC BEHAVIORS; PROBLEM ESTIMATION; RENORMALIZATION; STATISTICAL DATA ANALYSIS;

NONLINEAR SYSTEMS;

EID: 19944395196     PISSN: 01672789     EISSN: None     Source Type: Journal    
DOI: 10.1016/j.physd.2005.02.001     Document Type: Article

References (27)

1. e.g. M.E. Fisher, Scaling, Universality and Renormalization Group Theory, Lecture Notes in Physics, vol. 186, Springer, Berlin, 1986
2. M. Le Bellac Quantum and Statistical Field Theory 1991 Clarendon Press Oxford
3. Y. Oono Int. J. Mod. Phys. B 14 2000 1327
4. K. Nozaki, and Y. Oono Phys. Rev. E 68 2001 046101
5. Y. Shiwa Phys. Rev. E 63 2000 016119
6. This point of view including the idea of pursuit of stability was presented in the Cherry Bud Workshop at Yokohama held on March 2004; the paper is an outgrowth of the workshop presentation.
7. L.-Y. Chen, N. Goldenfeld, and Y. Oono Phys. Rev. E 54 1996 376
8. K. Hasselmann Tellus 28 1976 473
9. See, also L. Arnold, in: N.S. Namachchivaya, Y.K. Lin (Eds.), IUTAM on Nonlinear Stochastic Dynamics, Kluwer Academic Publishers, 2003, 5.
10. Y. Kifer Ergodic Theory and Dynamical Systems 24 2004 847
11. V. Bakhtin, and Y. Kifer Probab. Theory Relat. Fields 129 2004 157
12. This is only formal. 〈 f 〉 A may not even be continuous, and the reduced equation may not have any solution in some cases. The meaning of the approximation of the true solution x (t) in terms of A (t) is also quite delicate. Mathematically, we expect the modulus of the discrepancy integrated over [ 0, t / ε ] converges to zero as ε → 0
13. The interpretation of this stochastic equation is delicate. Our interpretation (conjecture) is, under the condition that the averaged equation (2.7) is well behaved, that ∫ d t (x ̇ - 〈 f 〉) 2 / 2 ε b (A) is the rate function for the large deviation of the true solution from, A (t). That is, we interpret the Langevin equation (2.11) as a shorthand notation of the large deviation, principle
14. For this point of view, see Oono Y. Prog. Theor. Phys. 89 1993 973.
15. G. Jona-Lasinio Nuovo Cimento B 26 1975 99
16. W. Feller, An Introduction to Probability Theory and Its Applications, vol. II, Wiley, New York, 1971, pp. 21-23.
17. M.D. Donsker An invariance principle for certain probability limit theorems Mem. Am. Math. Soc. 6 1951
18. e.g. P. Sprent, Data driven statistical methods, Chapman and Hall, 1998.
19. M. Paniconi Phys. Rev. E 57 1998 2690
20. G. Shafer, and V. Vovk Probability and Finance, It's Only a Game! 2001 Wiley New York
21. This statement is equivalent to asserting that probability theory is used to describe the possible world consistent with the given empirical data, because objective probability must be consistent with empirical frequencies. One might say that Bayesian or subjective interpretation of probability is possible and free from frequencies, so the above assertion is incomplete.
22. Here, the reader may wonder why we use only the inequality relations instead of more quantitative concepts such as distances to describe dissimilarities. It is our general belief that converting the dissimilarity information in terms of qualitative inequalities makes the analysis method much more robust against corrupted and/or missing data. Thus, our proposal is: converting all the quantitative data into a set of qualitative relations that could recover the quantitative relations if the original data is quantitative must be a general strategy in data analysis.
23. We know that the requirement is enough to recover a geometrical object if the number of points in the object is not too small (e.g., not less than 20).
24. T. Poggio, R. Rifkin, S. Mukherjee, and P. Niyoki Nature 428 2004 419
25. Y.-h. Taguchi, Y. Oono, Adv. Chem. Phys. 130B (2005) 315-351.
26. e.g. I. Borg, P. Groenen, Modern Multidimensional Scaling, Springer, New York, 1997.
27. Y.-h. Taguchi, Y. Oono, Bioinformatics, Advance Access published on February 22, 2005; doi:10.1093/bioinformatics/ bti067.