Abstract
We consider the generalization problem for a perceptron with binary synapses, implementing the Stochastic Belief-Propagation-Inspired (SBPI) learning algorithm which we proposed earlier, and perform a mean-field calculation to obtain a differential equation which describes the behaviour of the device in the limit of a large number of synapses N. We show that the solving time of SBPI is of order \(N\sqrt{\log N}\) , while the similar, well-known clipped perceptron (CP) algorithm does not converge to a solution at all in the time frame we considered. The analysis gives some insight into the ongoing process and shows that, in this context, the SBPI algorithm is equivalent to a new, simpler algorithm, which only differs from the CP algorithm by the addition of a stochastic, unsupervised meta-plastic reinforcement process, whose rate of application must be less than \(\sqrt{2/(\pi N)}\) for the learning to be achieved effectively. The analytical results are confirmed by simulations.
Similar content being viewed by others
References
Baldassi, C., Braunstein, A., Brunel, N., Zecchina, R.: Efficient supervised learning in networks with binary synapses. Proc. Natl. Acad. Sci. USA 104, 2079–2084 (2007)
Bhalla, U.S., Iyengar, R.: Emergent properties of networks of biological signaling pathways. Science 283, 381–387 (1999)
Bialek, W.: Stability and noise in biochemical switches. Adv. Neural Inf. Process. Syst. 13, 103–109 (2000)
Blum, A.L., Rivest, R.L.: Training a 3-node network is np-complete. Neural Netw. 5, 117–127 (1992)
Braunstein, A., Zecchina, R.: Learning by message-passing in networks of discrete synapses. Phys. Rev. Lett. 96, 030,201 (2006)
Engel, A., van den Broeck, C.: Statistical Mechanics of Learning. Cambridge University Press, Cambridge (2001)
Fusi, S., Drew, P.J., Abbott, L.F.: Cascade models of synaptically stored memories. Neuron 45(4), 599–611 (2005)
Golea, M., Marchand, M.: On learning perceptrons with binary weights. Neural Comput. 78, 333–342 (1993)
Gutfreund, H., Stein, Y.: Capacity of neural networks with discrete synaptic couplings. J. Phys. A Math. Gen. 23, 2613–2630 (1990)
Hayer, A., Bhalla, U.S.: Molecular switches at the synapse emerge from receptor and kinase traffic. PLoS Comput. Biol. 1(2), e20 (2005)
Kabashima, Y., Uda, S.: A BP-based algorithm for performing Bayesian inference in large perceptron-type networks. In: Algorithmic Learning Theory. Lecture Notes in Computer Science, vol. 3244, pp. 479–493. Springer, Berlin (2004)
Kinouchi, O., Caticha, N.: Optimal generalization in perceptrons. J. Phys. A 25, 6243 (1992)
Krauth, W., Mézard, M.: Storage capacity of memory networks with binary couplings. J. Phys. 50, 3057 (1989)
Miller, P., Zhabotinsky, A.M., Lisman, J.E., Wang, X.J.: The stability of a stochastic camkii switch: dependence on the number of enzyme molecules and protein turnover. PLoS Biology 3(4), e107 (2005)
O’Connor, D.H., Wittenberg, G.M., Wang, S.S.H.: Graded bidirectional synaptic plasticity is composed of switch-like unitary events. Proc. Natl. Acad. Sci. USA 102, 9679–9684 (2005)
Petersen, C.C., Malenka, R.C., Nicoll, R.A., Hopfield, J.J.: All-or-none potentiation at CA3-CA1 synapses. Proc. Natl. Acad. Sci. USA 95, 4732–4737 (1998)
Rosenblatt, F.: Principles of Neurodynamics. Spartan Books, New York (1962)
Rosen-Zvi, M.: On-line learning in the ising perceptron. J. Phys. A 33, 7277–7287 (2000)
Solla, S.A., Winther, O.: Optimal perceptron learning: an online Bayesian approach. In: On-Line Learning in Neural Networks. Cambridge University Press, Cambridge (1998)
Yedidia, J.S., Freeman, W.T., Weiss, Y.: Understanding belief propagation and its generalizations. In: Exploring Artificial Intelligence in the New Millennium, pp. 236–239. Morgan Kaufman, San Mateo (2003). Chap. 8
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Baldassi, C. Generalization Learning in a Perceptron with Binary Synapses. J Stat Phys 136, 902–916 (2009). https://doi.org/10.1007/s10955-009-9822-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-009-9822-1