fpga-neuron
stableAn educational neural network on an FPGA, from a single neuron to an XOR net.
lfpga add fpga-neuron
✓ Claimed by
@libfpga.
This listing is owner-controlled via a libfpga.yaml manifest
in the repository.
Source on GitHub → Docs / homepage
Verification
Signals earned by running the open toolchain against the public source. This is the registry's whole point: know it works before you clone it.
Last checked 2026-07-05 by the LibFPGA toolchain.
- 832 LUT4s, 0 flip-flops
About
An educational walkthrough: train a tiny MLP to solve XOR in dependency-free Python, quantize it to Q4.4 fixed point, and synthesize a fully combinational network (832 LUT4s, zero flip-flops) proven correct against a bit-exact software model on all 293 test vectors. Includes schematics, a training-loss plot, and float-vs-fixed decision-boundary heatmaps. The building blocks behind libfpga's neural micro-kit, shown by hand.