Abstract
The problem of learning by examples in ultrametric committee machines (UCMs) is studied within the framework of statistical mechanics. Using the replica formalism we calculate the average generalization error in UCMs with L hidden layers and for a large enough number of units. In most of the regimes studied we find that the generalization error, as a function of the number of examples presented, develops a discontinuous drop at a critical value of the load parameter. We also find that when L>1 a number of teacher networks with the same number of hidden layers and different overlaps induce learning processes with the same critical points.
Similar content being viewed by others
References
Variano, E., AMcCoy, J.H., Lipson, H.: Networks, dynamics, and modularity. Phys. Rev. Lett. 92, 188701 (2004)
Huerta, R., Rabinovich, M.: Reproducible sequence generation in random neural ensembles. Phys. Rev. Lett. 93, 238104 (2004)
Yoshioka, M.: Learning of spatiotemporal patterns in Ising-spin neural networks: analysis of storage capacity by path integral methods. Phys. Rev. Lett. 102, 158102 (2009)
Lin, I.-H., Wu, R.-K., Chen, C.-M.: Synchronization in a noise-driven developing neural network. Phys. Rev. E 84, 051923 (2011)
Saito, A., Taiji, M., Ikegami, T.: Inaccessibility in online learning of recurrent neural networks. Phys. Rev. Lett. 93, 168101 (2004)
Fiete, I.R., Seung, H.S.: Gradient learning in spiking neural networks by dynamic perturbation of conductances. Phys. Rev. Lett. 97, 048104 (2006)
Saito, H., Katahira, K., Okanoya, K., Okada, M.: Statistical mechanics of structural and temporal credit assignment effects on learning in neural networks. Phys. Rev. E 83, 051125 (2011)
Braunstein, A., Ramezanpour, A., Zecchina, R., Zhang, P.: Inference and learning in sparse systems with multiple states. Phys. Rev. E 83, 056114 (2011)
Koralek, A.C., Jin, X., Long, J.D. II, Costa, R.M., Carmena, J.M.: Corticostriatal plasticity is necessary for learning intentional neuroprosthetic skills. Nature 483, 331 (2012)
Bardin, J.: Neuroscience: Making connections. Nature 483, 394 (2012)
Mesgarani, N., Chang, E.F.: Selective cortical representation of attended speaker in multi-talker speech perception. Nature 485, 233 (2012)
Seung, H.S., Sompolinsky, H., Tishby, N.: Statistical mechanics of learning by examples. Phys. Rev. A 45, 6056–6091 (1992)
Neirotti, J.P.: Can a student learn optimally from two different teachers? J. Phys. A 43, 015101 (2010)
Neirotti, J.P.: Parallel strategy for optimal learning in perceptrons. J. Phys. A 43, 125101 (2010)
Neirotti, J.P., Franco, L.: Computational capabilities of multilayer committee machines. J. Phys. A 43, 445103 (2010)
Franco, L., Anthony, M.: The influence of oppositely classified examples on the generalization complexity of Boolean functions. IEEE Trans. Neural Netw. 17, 578–590 (2006)
Rammal, R., Toulouse, G., Virasoro, M.A.: Ultrametricity for physicists. Rev. Mod. Phys. 58, 765–788 (1986)
Engel, A., Köhler, H.M., Tschepke, F., Vollmayr, H., Zippelius, A.: Storage capacity and learning algorithms for two-layer neural networks. Phys. Rev. A 45, 7590–7609 (1992)
Monasson, R., O’Kane, D.: Domain of solutions and replica symmetry breaking in multilayer neural networks. Europhys. Lett. 27, 85–90 (1994)
Monasson, R., Zecchina, R.: Weight space structure and internal representations: a direct approach to learning and generalization in multilayer neural networks. Phys. Rev. Lett. 75, 2432–2435 (1995)
Schwarz, H.: Learning rule in a multilayer neural network. J. Phys. A 26, 5781–5794 (1993)
Acknowledgements
The author would like to acknowledge the friendly criticisms from Dr. Roberto C. Alamino and Dr Laura Rebollo-Neira.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Neirotti, J.P. Learning in Ultrametric Committee Machines. J Stat Phys 149, 887–897 (2012). https://doi.org/10.1007/s10955-012-0636-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-012-0636-1