fontinalis / fonet Goto Github PK

Examples and README examples need updating to fix Travis builds, and a few other function docs need changing in order to raise report card back to A+

Will be raising a PR for this

/*Geometric sequence
/
func PrepareTrainingData_dengbishulie(n int) [][][]float64 {
traindatas := make([][][]float64,0)
for i:=0;i<n;i++{
for j:=0;j<n;j++{
traindatas = append(traindatas,[][]float64{[]float64{float64(i),float64(j)},[]float64{float64(ij)}})
}
}
fmt.Printf("%++v",traindatas)
return traindatas
}

/main/
n,err := fonet.NewNetwork([]int{2, 1, 1},fonet.LeakyReLU)
if err != nil {
panic(err.Error())
}
trainingdata := PrepareTrainingData_dengbishulie(10)

n.Train(trainingdata,100,0.95,true)
fmt.Println(n.Predict([]float64{2,6}))

------output is NaN
why???

Please be aware of the following warning on Travis.org:

Please be aware travis-ci.org will be shutting down by end of May 2021. Please consider migrating to travis-ci.com.

I'm trying to work out if this is intended behaviour, whatever data I train a network with, the output is always between 0 and 1, even if that does not match the training data.

Please see #1 as further explanation.

Have I misunderstood the scope of the networks? Is this behaviour expected? If I needed a network which was able to output a float of arbitrary size, is this possible with this package?

Many thanks