- Access by Xinjiang University
The statistical mechanics of learning a rule
Rev. Mod. Phys. 65, 499 – Published 1 April, 1993
DOI: https://doi.org/10.1103/RevModPhys.65.499
Abstract
A summary is presented of the statistical mechanical theory of learning a rule with a neural network, a rapidly advancing area which is closely related to other inverse problems frequently encountered by physicists. By emphasizing the relationship between neural networks and strongly interacting physical systems, such as spin glasses, the authors show how learning theory has provided a workshop in which to develop new, exact analytical techniques.
References (213)
- Abu-Mostafa, Y.S., 1989, Neural Comput. 1, 312
- Ackley, D. H., G. E. Hinton, and T.J. Sejnowski, 1985, Cognitive Sci. 9, 147
- Amari, S.-I., and N. Murata, 1991, "Statistical theory of learning curves under entropic loss criterion," University of Tokyo preprint
- Amit, D. J., 1989, Modeling Brain Function (Cambridge University, Cambridge, England)
- Amit, D. J., H. Gutfreund, and H. Sompolinsky, 1985, Phys. Rev. Lett. 55, 1530
- Anderson, P. W., Phys. Today, "Reference Frame," Jan. 1988, March 1988, June, 1988, Sept. 1988, July 1989, Sept. 1989, and March 1990
- Anlauf, J. K., and M. Biehl, 1989, Europhys. Lett. 10, 687
- Anlauf, J. K., and P. Kuhlmann, 1992, "Metastable states in the projection rule network," Universität Giessen preprint
- Barkai, E., D. Hansel, and I. Kanter, 1990, Phys. Rev. Lett. 65, 2312
- Barkai, E., D. Hansel, and H. Sompolinsky, 1992, Phys. Rev. A 45, 4146
- Baum, E. B., 1991, IEEE Trans. Neural Networks 2, 5
- Baum, E. B., and D. Haussler, 1989, Neural Comput. 1, 151
- Baxter, R. J., 1982, Exactly Solved Models in Statistical Mechanics (Academic, London)
- Bex, G.-J., 1992, private communication
- Bichsel, M., and P. Seitz, 1989, Neural Networks 2, 133
- Biehl, M., 1989, Diplomarbeit (Universität Giessen)
- Biehl, M., J. K. Anlauf, and W. Kinzel, 1991, in Neurodynamics, edited by F. Pasemann and H. D. Doebner (World Scientific, Singapore), p. 194
- Biehl, M., and M. Opper, 1991, Phys. Rev. A 44, 6888
- Biehl, M., and H. Schwarze, 1992, Europhys. Lett. 20, 733
- Binder, K., and D. W. Heerman, 1988, Monte Carlo Simulations in Statistical Mechanics (Springer, Berlin)
- Binder, K., and A. P. Young, 1986, Rev. Mod. Phys. 58, 801
- Bohr, H., J. Bohr, S. Brunak, R. J. M. Cotterill, H. Fredholm, B. Lautrup, and S. B. Peterson, 1990, FEBS Lett. 261, 43
- Bollé, D., and P. Dupont, 1990, in Statistical Mechanics of Neural Networks, edited by L. Garrido (Springer, Berlin)
- Bollé, D., P. Dupont, and J. Huyghebaert, 1992, Phys. Rev. A 45, 4194
- Bollé, D., P. Dupont, and J. van Mourik, 1991, J. Phys. A 24, 1065
- Bronstein, I. N., and K. A. Semendjajew, 1985, Taschenbuch der Mathematik (Harri Deutsch, Thun), p. 677
- Brout, R., 1965, Phase Transitions (Benjamin, New York)
- Bryson, A. E., and Y.-C. Ho, 1969, Applied Optimal Control (Blaisdell, New York)
- Carnevali, P., and S. Patarnello, 1987, Europhys. Lett. 4, 1199
- Chesterton, G. K., 1910, "What's Wrong with the World," Chap. IV, p. 14
- Cottrell, G. W., P. Munro, and D. Zipser, 1987, in 9th Annual Conference of the Cognitive Science Society (Erlbaum, Hillsdale, NJ), p. 462
- Cover, T.M., 1965, IEEE Trans. Electron. Comput. EC-14, 326
- de Almeida, J.R.L., and D. J.Thouless, 1978, J. Phys. A. Math. Nucl. Gen. 11, 983
- De Dominicis, C., 1978, Phys. Rev. B 18, 4913
- De Figueiredo, R.J. P., 1980, J. Math. Anal. Appl. 38, 1227
- Del Giudice, P., S., Franz, and M. A. Virasoro, 1989, J. Phys. (Paris) 50, 121
- Denker, J. S., D. Schwartz, B. Wittner, S. A. Solla, R. E.Howard, L. D. Jackel, and J. J. Hopfield, 1987, Complex Syst. 1, 877
- Derrida, B., 1981, Phys. Rev. B 24, 2613
- Derrida, B., E. Gardner, and A. Zippelius, 1987, Europhys. Lett. 4, 167
- Derrida, B., R. B. Griffiths, and A. Prügel-Bennett, 1991, J. Phys. A 24, 4907
- Derrida, B., V. Hakim, and J. Vannimenus, 1991, Phys. Rev. Lett. 43, 888
- Derrida, B., and J. P. Nadal, 1987, J. Stat. Phys. 49, 993
- Diederich, S., and M. Opper, 1987, Phys. Rev. Lett. 58, 949
- Domany, E., W. Kinzel, and R. Meir, 1986, Europhys. Lett. 2, 175
- Domany, E.W. KinzelR. MeirDomany, E., J. L. van Hemmen, and K. Schulten, 1991, Eds. Models of Neural Networks (Springer, Berlin)
- Eckmiller, R., 1989, in Neural Computing Architectures. The Design of Brainlike Machines, edited by I. Aleksander (North Oxford Academic, London), p. 305
- Edwards, S. F., 1970, in 4th International Conference on Amorphous Materials, edited by R. W. Douglas and B. Ellis (Wiley, New York)
- Edwards, S. F., and P. W. Anderson, 1975, J. Phys. F 5, 965
- Edwards, S. F., and R. C. Jones, 1976, J. Phys. A 9, 1595
- Eisenstein, E., and I. Kanter, 1992, "Numerical study of backpropagation learning algorithms for multilayer networks," Bar-Ilan University preprint
- Engel, A., H. M. Köhler, F. Tschepke, H. Vollmayr, and A. Zippelius, 1992, Phys. Rev. A 45, 7590
- Fahlmann, S. E., 1989, in Proceedings of the 1988 Connectionist Models Summer School, edited by D. Touretzky, G. Hinton, and T. Sejnowski (Morgan Kaufmann, San Mateo), p. 38
- Fahlman, S. E., and C. Lebiere, 1990, in Advances in Neural Information Processing Systems II, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo), p. 524
- Falk, H., 1975, J. Phys. C 8, L294
- Fisher, M. E., and D. S. Gaunt, 1964, Phys. Rev. A 133, 224
- Frean, M., 1990, Neural Comput. 2, 198
- Gallant, S. I., 1986, in Proceedings of the Eighth Annual Conference of the Cognitive Science Society (Erlbaum, Hillsdale, NJ), p. 652
- Gardner, E., 1986, J. Phys. A 19, L1047
- Gardner, E., 1987, Europhys. Lett. 4, 481
- Gardner, E., 1988, J. Phys. A 21, 257
- Gardner, E., and B. Derrida, 1988, J. Phys. A 21, 271
- Gardner, E., and B. Derrida, 1989, J. Phys. A 22, 1983
- Gardner, E., I. Yekutieli, and H. Gutfreund, 1989, J. Phys. A 22, 1995
- Garey, M. R., and D. S. Johnson, 1979, Computers and Intractability: A Guide to NP-Completeness (Freeman, New York)
- Geszti, T., 1990, Physical Models of Neural Networks (World Scientific, Singapore)
- Golea, M., and M. Marchand, 1990, Europhys. Lett. 12, 205
- Gross, D. J., and M. Mézard, 1984, Nucl. Phys. B 240, 431
- Grossman, T., 1990, in Advances in Neural Information Processing Systems II, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo), p. 516
- Grossman, T., and E. Domany, 1992, private communication
- Grossman, T., R. Meir, and E. Domany, 1988, Complex Syst. 2, 555
- Gutfreund, H., and G. Toulouse, 1992, "The physics of neural networks," in Spin Glasses and Biology, edited by D. L. Stein (World Scientific, Singapore)
- Györgyi, G., 1990a, Phys. Rev. Lett. 64, 2957
- Györgyi, G., 1990b, Phys. Rev. A 41, 7097
- Györgyi, G., and N. Tishby, 1990, in Neural Networks and Spin Glasses, edited by W. K. Theumann and R. Köberle (World Scientific, Singapore), p. 3
- Hammann, J., M. Ocio, E. Vincent, and R. Orbach, 1992, Physica A 185, 278
- Hansel, D., G. Mato, and C. Meunier, 1992, Europhys. Lett. 20, 471
- Hansel, D., 1992, private communication
- Hansel, D., and H. Sompolinsky, 1990, Europhys. Lett. 11, 687
- Hanson, S. J., and L. Pratt, 1989, in Advances in Neural Information Processing Systems I, edited by D.S. Touretzky (Morgan Kaufmann, San Mateo), p. 117
- Hebb, D. O., 1949, The Organization of Behavior (Wiley, New York)
- Hecht-Nielsen, R., 1987, in IEEE 1st International Conference on Neural Networks, Vol. III, edited by M. Caudill and C. Butler (IEEE, New York). p. 455
- Helmbold, D. P., and P.M. Long, 1991, in Proceedings of the Fourth Annual Workshop on Computational Learning Theory, edited by M. K. Warmuth and L. G. Valliant (Morgan Kaufmann, San Mateo)
- Hertz, J. A., A. Krogh, and R. G. Palmer, 1991, Introduction to the Theory of Neural Computation (Addison-Wesley, Redwood City, CA)
- Hertz, J. A., A.Krogh, and G. I. Thorbergsson, 1989, J. Phys. A 22, 2133
- Hinton, G. E., 1986, in Proceedings of the 8th Annual Conference of the Cognitive Science Society (Erlbaum, Hillsdale, NJ), p. 1
- Hopfield, J. J., 1982, Proc. Natl. Acad. Sci. U.S.A. 79, 2554
- Horner, H., 1992a, Z. Phys. B 86, 291
- Horner, H., 1992b, Z. Phys. B 87, 371
- Huang, K., 1987, Statistical Mechanics (Wiley, New York), p. 427
- Ising, E., 1925, Z. Phys. 31, 253
- Itzykson, C., and J.-M. Drouffe, 1989, Statistical Field Theory II (Cambridge University, Cambridge, England)
- Jacobs, R. A., 1988, Neural Networks 1, 295
- Jaynes, E. T., 1957, Phys. Rev. 106, 620
- Jaynes, E.T., 1983, Papers on Probability, Statistics and Statistical Physics, Synthese Library Vol. 158, edited by R. D. Rosenkranz (Reidel, Dordrecht)
- Jeffreys, H., 1939, Theory of Probability (Clarendon, Oxford)
- Kac, M., 1968, in Trondheim Theoretical Physics Seminar, Nordita Publ. No. 286
- Kanter, I., 1988, Phys. Rev. A 37, 2739
- Kanter, I., 1992, private communication
- Karl, J. H., 1989, An Introduction to Digital Signal Processing (Academic, San Diego)
- Kinouchi, O., and N. Caticha, 1992, "Optimal generalization in perceptrons," Universade de São Paulo preprint
- Kinzel, W., and M. Opper, 1991, in Models of Neural Networks, edited by E. Domany, J. L. van Hemmen, and K. Schulten (Springer, Berlin), p. 149
- Kinzel, W,. and P. Ruján, 1990, Europhys. Lett. 13, 473
- Knerr, S., L. Personnaz, and G. Dreyfus, 1990, in Neurocomputing: Algorithms, Architectures and Applications (NATO ASI Series) (Springer, Berlin)
- Kocher, I., and R.Monasson, 1991, Int. J. of Neural Syst. 2, 115
- Köhler, H., S. Diederich, W. Kinzel, and M. Opper, 1990, Z. Phys. B 78, 333
- Kohonen, T., 1988, Neural Network Architectures (Kogan Page, London)
- Kramer, A. H., and A. Sangiovanni-Vincentelli, 1989, in Advances in Neural Information Processing Systems I, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo)
- Krauth, W., and M. Mézard, 1987, J. Phys. A 20, L745
- Krauth, W., and M. Mézard, 1989, J. Phys. (Paris) 50, 3057
- Krogh, A., 1992, J. Phys. A 25, 1119
- Krogh, A., and J. A. Hertz, 1991, in Advances in Neural Information Processing Systems III, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo)
- Krogh, A., G.I. Thorbergsson, and J. A. Hertz, 1990, in Advances in Neural Information Processing Systems II, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo), p. 733
- Kühn, R., S. Bös, and J. L. van Hemmen, 1991, Phys. Rev. A 43, 2084
- Kurchan, J., and E. Domany, 1990, J. Phys. A 23, L847
- Langevin, P., 1908, Compt. Rend. 43, 530
- Le Cun, Y., 1986, in Disordered Systems and Biological Organization, edited by E. Bienenstock, F. Fogelman-Soulié, and G. Weisbuch (Springer, Berlin), p. 234
- Le Cun, Y., J. S. Denker, and S. A. Solla, 1990, in Advances in Neural Information Processing Systems II, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo), p. 598
- Le Cun, Y., I. Kanter, and S. A.Solla, 1991, Phys. Rev. Lett. 66, 2396
- Levin, E., N. Tishby, and S. A. Solla, 1989, in Proceedings of the 2nd Workshop on Computational Learning Theory (Morgan Kaufmann, San Mateo)
- Levin, E., Y. Le Cun, and V. Vapnik, 1992, "Measuring the capacity of a learning machine (II)," AT&T Bell Laboratories, preprint
- Lippman, R. P., 1989, Neural Comput. 1, 1
- MacKay, D. J. C., 1992, Neural Comput. 4, 448
- Makram-Ebeid, S., J.-A. Sirat, and J.-R. Viala, 1989, in International Joint Conference on Neural Networks (IEEE, New York), Vol. II, p. 373
- Marchand, M., M. Golea, and P. Ruján, 1990, Europhys. Lett. 11, 487
- Martinez, D., and D. Esteve, 1992, Europhys. Lett. 18, 95
- Mato, G., and N. Parga, 1992, J. Phys. A 25, 5067
- McCulloch, W. S., and W. Pitts, 1943, Bull. Math. Biophys. 5, 115
- Meir, R., and J. F. Fontanari, 1992, Phys. Rev. A 45, 8874
- Mézard, M., and J.-P. Nadal, 1989, J. Phys. A 22, 2191
- Mézard, M., J.-P. Nadal, and G. Toulouse, 1986, J. Phys. (France) 47, 1457
- Mézard, M., G. Parisi, and M. A. Virasoro, 1987, Spin Glass Theory and Beyond (World Scientific, Singapore)
- Mézard, M., and S. Patarnello, 1989, "On the capacity of feedforward layered networks," Ecole Normale Supérieure Preprint No. LPTENS 89/24
- Minsky, M. L., and S. A. Papert, 1969, Perceptrons (MIT, Cambridge, MA)
- Mitchell, T. M., 1982, Artif. Intell. 18, 203
- Mitchison, G. J., and R. M. Durbin, 1989, Biol. Cybern. 60, 345
- Moody, J., 1992, "Generalization, weight decay and architecture selection for nonlinear learning schemes," Yale University preprint
- Müller, B., and J. Reinhardt, 1991, Neural Networks—An Introduction (Springer, Berlin)
- Nabutovsky, D., and E. Domany, 1991, Neural Comput. 3, 604
- Nabutovsky, D., T. Grossman, and E. Domany, 1990, Complex Syst. 4, 519
- Nadal, J.-P. 1989, Int. J. Neural Syst. 1, 55
- Nadal, J.-P., and A. Rau, 1991, J. Phys. I France 1, 1109
- Onsager, L., 1944, Phys. Rev. 65, 117
- Opper, M., 1989, Europhys. Lett. 8, 389
- Opper, M., S. Diederich, and J. K. Anlauf, 1988, in Neural Networks from Models to Applications, edited by L. Personnaz and G. Dreyfus (I.D.S.E.T., Paris)
- Opper, M., and D. Haussler, 1991a, Phys. Rev. Lett. 66, 2677
- Opper, M., and D. Haussler, 1991b, in Proceedings of the 4th Annual Workshop on Computational Learning Theory, edited by M. K. Warmuth and L. G. Vallian (Morgan Kaufmann, San Mateo)
- Opper, M., W. Kinzel, J. Kleinz, and R. Nehl, 1990, J. Phys. A 23, L581
- Parisi, G., 1988, Statistical Field Theory (Addison-Wesley, Redwood City, CA)
- Parisi, G., and F. Slanina, 1992, Europhys. Lett. 17, 497
- Parrando, J. H. R., and C. Van den Broeck, 1992, "Vapnik Chervonenkis bounds for generalization," University of California, San Diego, preprint
- Patel, H. K., 1992, "Computational complexity, learning, rules and storage capacities: A Monte Carlo study for the binary perceptron," Universität Heidelberg preprint
- Patarnello, S., and P. Carnevali, 1987, Europhys. Lett. 4, 503
- Penney, R. W., and A. Wendenmuth, 1992, private communication
- Peretto, P., 1984, Biol. Cybern. 50, 51
- Peretto, P., 1991, The Modeling of Neural Networks (Cambridge University, Cambridge, England)
- Personnaz, L., I. Guyon, and G. Dreyfus, 1986, Phys. Rev. A 34, 4217
- Plaut, D., S. Nowlan, and G. Hinton, 1986, "Experiments on learning by back propagation," Technical Report No. CMU-CS-86-126 (Department of Computer Science, Carnegie-Mellon University, Pittsburgh, PA)
- Ritter, H., K. Schulten, and T. Martinetz, 1990, Neuronale Netze (Addison-Wesley, Bonn)
- Rohwer, R., 1990, in Advances in Neural Information Processing Systems II, edited by D. S. Touretzky (Morgan Kaufmann, San Mateo), p. 558
- Rosenblatt, F., 1962, Principles of Neurodynamics (Spartan, New York)
- Ruján, P., and M. Merchand, 1989, Complex Syst. 3, 229
- Ruján, P.M. MerchandRumelhart, D. E., and J. L. McClelland, 1986, Eds., Parallel Distributed Processing (MIT, Cambridge, MA)
- Saad, D., and E. Marom, 1990, Complex Syst. 4, 107
- Schmitz, H. J., G. Pöppel, F. Wünsch, and U. Krey, 1990, J. Phys. (France) 51, 167
- Schwartz, D. B., V. K. Samalam, S. A. Solla, and J. S. Denker, 1990, Neural Comput. 2, 371
- Schwarze, H., 1991, Diplomarbeit (Universität Giessen)
- Schwarze, H. and J. Hertz, 1992a, Europhys. Lett. 20, 375
- Schwarze, H., and J. Hertz, 1992b, "Generalization in a fully-connected committee machine," Nordita Preprint No. 92/61 S
- Schwarze, H., M. Opper, and W. Kinzel, 1992, Phys. Rev. A 46, R6185
- Sejnowski, T. J., and C. R. Rosenberg, 1987, Complex Syst. 1, 145
- Seung, H. S., 1991, private communication
- Seung, H. S., M. Opper, and H. Sompolinsky, 1992, "Query by committee," Proceedings of the 5th Annual ACM Workshop on Computational Learning Theory (ACM, NY)
- Seung, H. S., H. Sompolinsky, and N. Tishby, 1992, Phys. Rev. A 45, 6056
- Sherrington, D., 1986, Prog. Theor. Phys. 87, 180
- Sherrington, D., and S. Kirkpatrick, 1975, Phys. Rev. Lett. 35, 1792
- Sietsma, J., and R. J. F. Dow, 1988, in IEEE International Joint Conference on Neural Networks, Vol. I (IEEE, New York), p. 325
- Sirat, J.-A., and J.-P. Nadal, 1990, Network 1, 423
- Sjöberg, L. Ljung, 1992, "Overtraining, regularization and searching for minimum in neural networks," Linköping University, Sweden, preprint
- Solla, S. A., 1992, private communication
- Solla, S. A., E. Levin, and M. Fleischer, 1988, Complex Syst. 2, 625
- Sompolinsky, H., and N. Tishby, 1990, Europhys. Lett. 13, 567
- Sompolinsky, H., N. Tishby, and H. S. Seung, 1990, Phys. Rev. Lett. 65, 1683
- Sourlas, N., 1989, Nature 339, 693
- Stanley, H. E., 1971, Introduction to Phase Transitions and Critical Phenomena (Oxford University Press, Oxford)
- Tishby, N., E. Levin, and S. Solla, 1989, in IEEE International Joint Conference on Neural Networks, Vol. II (IEEE, New York), p. 403
- Utans, J., and J. Moody, 1991, in Proceedings of the First International Conference on Artificial Intelligence Applications on Wall Street (IEEE Computer Society, Los Alamos, CA)
- Vallet, F., 1989, Europhys. Lett. 8, 747
- Vallet, F., J. Cailton, and P. Refregier, 1989, Europhys. Lett. 9, 315
- Van den Broeck, C., and R. Kawai, 1990, Phys. Rev. A 42, 6210
- van Hemmen, J. L., D. Grensing, A. Huber, and R. Kühn, 1988, J. Stat. Phys. 50, 231
- van Hemmen, J. L., G. Keller, and R. Kühn, 1988, Europhys. Lett. 5, 663
- Vapnik, V. N., 1982, Estimation of Dependences Based on Empirical Data (Springer, Berlin)
- Vapnik, V. N., 1992, "Measuring the capacity of a learning machine (I)," AT&T Bell Laboratories preprint
- Vapnik, V. N., and A. Y. Chervonenkis, 1971, Theory Probab. Its Appl. 16, 264
- Venkatesh, S., 1986, in Neural Networks for Computing, AIP Conference Proceedings No. 151, edited by J. S. Denker (AIP, New York), p. 440
- Vogl, T. P., W. T. Zink, J. K. Mangis, D. L. Alkon, and A. K. Rigler, 1988, Biol. Cybern. 59, 257
- von Lehmann, A., E. G. Paek, P. F. Iao, A. Marrakchi, and J. S. Patel, 1988, in IEEE International Conference on Neural Networks, Vol. I (IEEE, New York), p. 335
- Waibel, A., T. Hanazawa, G. Hinton, K. Shikano, and K. J. Lang, 1989, IEEE Trans. Acoust. Speech Signal Process. 37, 328
- Watkin, T. L. H., 1991, unpublished
- Watkin, T. L. H., 1993, "Optimal learning with a neural network," Europhys. Lett. 21, 871
- Watkin, T. L. H., and A. Rau, 1992a, J. Phys. A 25, 113
- Watkin, T. L. H., and A. Rau, 1992b, Phys. Rev. A 45, 4102
- Watkin, T. L. H., A. Rau, D. Bollé, and J. van Mourik, 1992, J. Phys. I France 2, 167
- Watkin, T. L. H., and D. Sherrington, 1992, "An exactly solved neural network which is susceptible to simulation," Oxford University preprint
- Watrous, R. L., 1987, in IEEE 1st International Conference on Neural Networks, edited by M. Caudill and C. Butler (IEEE, New York), Vol. II, p. 619
- Weisbuch, G., 1991, Complex Systems Dynamics (Addison-Wesley, Redwood City, CA)
- Wendemuth, A., 1991, Diplomarbeit (Universität Giessen)
- Widrow, B., and M. E. Hoff, 1960, in 1960 IRE WESCON Conv. Rep. IV (Institute of Radio Engineers, New York), p. 96
- Winder, R. O., 1963, IEEE Trans. Electron. Comput. EC-12, 561
- Wittner, B. S., and J. S. Denker, 1988, in Neural Information Processing Systems, edited by D. Z. Anderson (AIP, New York), p. 850
- Wong, K. Y. M., and D. Sherrington, 1990, J. Phys. A 23, 4659
- Wu, F. Y., 1982, Rev. Mod. Phys. 54, 235
- Zollner, R. H. J. Schmitz, F. Wünsch, and U. Krey, 1992, Neural Networks 5, 771