HLM algorithm-extension to pattern recognition

Neural networks have allowed the construction of various tools adapted to the resolution of very different problems like automatic classification and pattern recognition. The HLM algorithm which serves for classification may be seen as a particular case of pattern recognition, since it doesn´t permit the recognition of position, orientation and shape of a given object. The authors propose two extensions of the HLM algorithm (HLM1, HLM2) which make use of the recognition of noisy patterns. This is the topic that the authors tackle in the paper. The paper includes the processing of the algorithms following a local interconnection scheme. The authors also present convergence techniques of these algorithms by backpropagation of the desired outputs