In this paper, we propose a hyper-spherical surface interactive interc
onnections (HSII) learning method for category learning. This method g
enerates the number of hidden units based on the same algorithm as a r
estricted coulomb energy (RCE) network and updates weights and biases
by the error backpropagation algorithm (EBP). The HSII method adapts d
esired features of RCE and EBP methods. The HSII (1) optimizes the net
work size, (2) learns samples quickly, and (3) has well generalized pe
rformance on untrained inputs. We demonstrate application of the HSII
to recognition of characters on X-ray films. We also compared performa
nces of the HSII with the RCE using receiver operating characteristic
analysis.