Name		Name	Last commit message	Last commit date
Dataset.hpp		Dataset.hpp
FCNet.hpp		FCNet.hpp
Makefile		Makefile
Matrix.hpp		Matrix.hpp
README.md		README.md
SGDTrainer.hpp		SGDTrainer.hpp
core.hpp		core.hpp
eval.hpp		eval.hpp
main.cpp		main.cpp

Repository files navigation

uNet

Minimalistic fully connected neural network library.

Run training on MNIST

make download
make compile
./main

Examples

Accuracy on MNIST: 97.09%
- Hidden layers: 400-50 ReLu activation
Accuracy on MNIST: 98.08%
- Hidden layers: 800 Sigmoid activation

About

Minimalistic NN framework implementation

1 star

1 watching

Report repository

Releases

Packages

Contributors

Languages