A self-organizing neural network for classifying sequences

The ability to recognize sequences is important for applications such as speech processing, vision, and control systems. A self-organizing neural network model that is able to form an ordered map of a sequence is presented. The model is based on extensions to T. Kohonen's self-organizing topology maps (Self-Organization and Associative Memory, Springer-Verlag, 1984). Theoretical results and simulations are presented that demonstrate the ability of the model to learn arbitrary sequences of n-dimensional patterns. The network model represents a learned sequence with a fixed sequence of network outputs that is easily identifiable. This representation makes the development of a sequence classifier relatively simple.<<ETX>>