Tiny NPU

Description:

This Tiny Tapeout project implements a fixed-weight two-layer binary MLP. It evaluates eight hidden neurons serially over the input vector, then runs a second serial output layer to select a 2-bit class and confidence score.