A small C++ neural network that trains on MNIST, reports training and test
accuracy, and saves its dense-layer weights under models/example/.
From the project root:
make trainTo build the drawing prediction app, SFML must be available on your system. On macOS, the simplest setup is:
brew install sfmlmake draw_predictThen run:
make run-drawUse the on-screen buttons, press Enter to predict, press C to clear, or press Escape to quit.
Keep the MNIST files in data/, then run:
./trainThe trained parameters are written to models/example/layer1.txt and
models/example/layer2.txt.
Application files are separate entry points. Do not compile train.cpp and
draw_predict.cpp into the same executable.