Shin'ichi Tamura
ON INTERPRETATIONS OF A FEED-FORWARD
NEURAL NETWORK
Abstract:In this technical report, interpretations of a feed-forward neural
network are described. For the interpretations, a network is decomposed into a
successive combination of unit transformations. A unit transformation stands for
a transformation from one layer output to the next higher layer output. This
transformation is further divided into an affine part and a sigmoid non-linear
part. The affine part is interpreted using the singular value decomposition
technique and the sigmoid non-linear part is interpreted based on a classification
of its input space into direction-invariant sub-space. Also described is an
application of the interpretations, which is applied to an exclusive OR feed-forward neural network.