An on-line learning rule, based on the introduction of a matrix momentum te
rm, is presented, aimed at alleviating the computational costs of standard
natural gradient learning. The new rule, natural gradient matrix momentum,
is analysed in the case of two-layer feed-forward neural network learning v
ia methods of statistical physics. It appears to provide a practical algori
thm that performs as well as standard natural gradient descent in both the
transient and asymptotic regimes but with a hugely reduced complexity.