The auto-associative neural network - a network architecture worth considering

Of the over 73,000 papers mentioning neural networks in the last 10 years, only 232 of them mention the auto-associative neural network (AANN). This is surprising, in that the AANN is a particularly useful architecture able to perform filtering, system modeling, anomaly detection as well as its apparently more traditional associative memory role. The purpose of this paper is to introduce this versatile network to a wider audience and to raise the awareness of the engineering community to a useful tool. A brief introduction to the architecture is followed with a brief discussion of some theoretical issues, and then moves on to few practical suggestions to get the interested practitioner started using them effectively.