A polynomial training algorithm for calculating perceptrons of optimal stability
Jorg Imhoff
Abstract
Recomi (REpeated COrrelation Matrix Inversion) is a polynomially fast algorithm for searching optimally stable solutions of the perceptron learning problem. For random unbiased and biased patterns it is shown that the algorithm is able to find optimal solutions, if any exist, in at worst O(N4) floating point operations. Even beyond the critical storage capacity alphac the algorithm is able to find locally stable solutions (with negative stability) at the same speed. There are no divergent time scales in the learning process. A full proof of convergence cannot yet be given, only major constituents of a proof are shown.
Create a lesson
Related papers
Knots in Condensed Matters
Y. M. Cho
Bouchaud's model exhibits two different aging regimes in dimension one
Gerard Ben Arous, Jiri Cerny
Periodic diffraction patterns for 1D quasicrystals
Pawel Buczek, Lorenzo Sadun, Janusz Wolny
Adiabatic association of ultracold molecules via magnetic field tunable interactions
Krzysztof Goral, Thorsten Koehler, Simon A. Gardiner et al.
High-Temperature Atomic Superfluidity in Lattice Boson-Fermion Mixtures
F. Illuminati, A. Albus
Constructive Methods of Invariant Manifolds for Kinetic Problems
A. N. Gorban, I. V. Karlin, A. Yu. Zinovyev