FPGA Neural Network with Online Training

This project is a two-layer neural network with no hidden layers implemented on an FPGA. The operation of this design is as follows:

Press a button assigned to either train or test.
If training is pressed and released, a full forward and backward propagation occurs on the image currently in memory (input by a .coe file to the BRAM or through other means).
If testing is pressed and released, only forward propagation occurs.
In both cases, a number representing the label of the image currently in memory is output, and the neural network's prediction for that image is output to the hex driver next to it.

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
HexDriver.sv		HexDriver.sv
LICENSE		LICENSE
README.md		README.md
Synchronizers.sv		Synchronizers.sv
calc_prediction.sv		calc_prediction.sv
control.sv		control.sv
forprop_neuron.sv		forprop_neuron.sv
image_ram.sv		image_ram.sv
labels_ram.sv		labels_ram.sv
multiplier16.sv		multiplier16.sv
multiplier24.sv		multiplier24.sv
multiplier8.sv		multiplier8.sv
negedge_detector.sv		negedge_detector.sv
network.sv		network.sv
nn_top.sv		nn_top.sv
sigmoid_func.sv		sigmoid_func.sv
weights_ram.sv		weights_ram.sv

Provide feedback