neural-net-go

README

Neural Net Go

Simple neural network implementation in Go.

Build

go build

Help

./neural-net-go -help to display all available flags.

Run the Iris sample

Dataset included.

./neural-net-go -model=models/iris.1.model -preset=iris -action=train
./neural-net-go -model=models/iris.1.model -preset=iris -action=test

Run the MNIST sample

Download the MNIST training and test data from https://pjreddie.com/projects/mnist-in-csv/ and place in the datasets directory.

./neural-net-go -model=models/mnist.1.model preset=mnist -action=train 
./neural-net-go -model=models/mnist.1.model preset=mnist -action=test 

After training, the model is saved to a JSON file. You can load the same model to train additional epochs or test its accuracy.

References

Thanks to:

• Daniel Whitenack for Building a neural network from scratch in Go
• Chang Sau Sheong for How to build a simple artificial neural network with Go