Maximum-throughput implementation of piecewise-linear fuzzy systems
Riccardo Rovatti, Michele Borgatti · 2002
A novel fuzzy-and is introduced for n-input fuzzy systems which reduces the cost of an inference evaluation from O(const/sup n/) to O(nlogn). Based on this connective, the architecture of a dedicated digital circuit is also sketched, which is able to compute one fuzzy inference every n+1 memory read cycles. This throughput is maximum as not less than n+1 independent piece of information are needed to infer a point on a n-dimensional surface.