ar X iv : c on d - m at / 0 20 23 50 20 F eb 2 00 2 Cryptography based on neural networks - analytical results
نویسنده
چکیده
Mutual learning process between two parity feed-forward networks with discrete and continuous weights is studied analytically, and we find that the number of steps required to achieve full synchronization between the two networks in the case of discrete weights is finite. The synchronization process is shown to be non-self-averaging and the analytical solution is based on random auxiliary variables. The learning time of an attacker that is trying to imitate one of the networks is examined analytically and is found to be much longer than the synchronization time. Analytical results are found to be in agreement with simulations.
منابع مشابه
ar X iv : m at h - ph / 0 60 20 16 v 1 7 F eb 2 00 6 Magnetic Geodesic Flows on Coadjoint Orbits ∗
We describe a class of completely integrable G-invariant magnetic geodesic flows on (co)adjoint orbits of a compact connected Lie group G with magnetic field given by the Kirillov-Konstant 2-form.
متن کاملar X iv : 0 81 1 . 23 46 v 3 [ m at h . A G ] 4 F eb 2 00 9 EFFECTIVE ALGEBRAIC DEGENERACY
We show that for every generic smooth projective hypersurface X ⊂ P, n > 2, there exists a proper algebraic subvariety Y $ X such that every nonconstant entire holomorphic curve f : C → X has image f(C) which lies in Y , provided degX > n n+5 . Table of contents
متن کاملar X iv : n uc l - th / 0 20 20 15 v 1 5 F eb 2 00 2 Toward a global description of the nucleus - nucleus interaction
Extensive systematizations of theoretical and experimental nuclear densities and of optical potential strengths extracted from heavy-ion elastic scattering data analyses at low and intermediate energies are presented. The energy-dependence of the nuclear potential is accounted for within a model based on the nonlocal nature of the interaction. The systematics indicates that the heavy-ion nuclea...
متن کامل