Symmetries in neural networks: a linear group action approach
R. Folk, A Kartashov, Petr Lisoněk, Peter Paule · Journal of Physics A Mathematical and General · 1993
Neural networks storing Hadamard patterns have been completely classified with respect to permutation symmetry. The symmetry group of the Hadamard patterns is found to be isomorphic to GL(n,F 2 ), and the symmetry groups of the networks are explicitly constructed for the most important classes. The volumes of different equivalence classes have been calculated.