engleski

Memory reduction techniques for holographic neurons

This paper shows how certain numerical transformations can be used to control the memory of a holographic neuron. More precisely, the techniques are considered that limit the magnitude or number of entries in the memory matrix, thus reducing the memory persistence and size respectively. Possible applications include short-term forecasting and data compression. Some illustrative experimental results are presented.

artificial neural networks; holographic neural technology; memory reduction; forecasting; data compression

nije evidentirano