Implementation issues for on-chip learning with analogue VLSI MLPS
Cairns GA., Tarassenko L.
Some of the issues concerning in-situ learning with analogue VLSI multi-layer perceptron (MLP) networks are addressed. The modes used to train analogue neural networks are considered, weight storage and circuit precision issues are studied, and the most promising training algorithms are identified. Some conclusion are made based on results from analogue VLSI chips that have been designed, built and successfully tested.