Preprocessing of the input vectors for the linear associator neural networks

This paper presents a methodology for ensuring the input to the linear associator neural network to be all linearly independent. This is a required condition for the linear associator neural network in order to produce exact output during recall. A number of linear associators may be connected in parallel to increase the capacity. A method to group all the input vectors as a set of linearly independent vectors is presented. The performance of the model is discussed. >