Build and run using:
$ go build .
$ ./godeep
Model setup looks like
model := godeep.Sequential{}
model.Init(len(trainingInputs[0]),"CROSSENTROPY")
// Create new layer and add to model.
layer1 := layers.BaseLayer{}
layer1.Init(40)
model.Add( &layer1 )
// Create new layer and add to model.
layer2 := layers.BaseLayer{}
layer2.Init(20)
model.Add( &layer2 )
// Create new layer and add to model.
layer3 := layers.BaseLayer{}
layer3.Init(10)
model.Add( &layer3 )
model.Compile()
model.Train(trainingInputs,trainingTargets,10,20)
Regularisation, Dropout and alternative activation functions (to sigmoidal) have not been implemented yet.