The generalized AdaTron algorithm
P. Nachbar, Josef A. Nossek, Josef Strobl · 1993 IEEE International Symposium on Circuits and Systems · 2002
The AdaTron algorithm can find the most insensitive weights of a perceptron or attractor network with respect to the usual Euclidean norm. The authors define various notions of robustness for q-norms as appropriate for perceptrons or attractor networks. By extending the so-called AdaTron theorem, they are able to generalize the AdaTron algorithm, which then finds the most insensitive weights with respect to an arbitrary q-norm.>