Frequency-based model of memory in neural networks

A neural-network model is presented that uses the frequency content of a signal to store and process information. The model is described theoretically and compared to amplitude-based neural networks. A version of the network is also simulated and its storage performance evaluated. A theoretical motivation for such a model is derived from M.L. Minsky's (1986) knowledge-line (k-line) theory of memory. The concept of a k-line is interpreted and associated with a frequency supportive network and a practical model for the implementation of a k-line-based memory is proposed. The implementation of a frequency-based network as a pattern recognizer is considered, along with possible specific applications.<
>