A Silicon Primitive for Competitive Learning
David Hsu, Miguel E. Figueroa, Chris Diorio · 2000
Competitive learning is a technique for training classification and clustering networks. We have designed and fabricated an l 1transistor primitive, that we term an automaximizing bump circuit, that implements competitive learning dynamics. The circuit performs a similarity computation, affords nonvolatile storage, and implements simultaneous local adaptation and computation. We show that our primitive is suitable for implementing competitive learning in VLSI, and demonstrate its effectiveness in a standard clustering task.