func DSoftmax(X *Matrix.Matrix) *Matrix.Matrix {
Total := 1 / X.TaxicabNorm()
Y := X.Scalar(complex(Total, 0))
S, _ := Matrix.Sustract(Matrix.FixValueMatrix(X.GetNColumns(), X.GetNColumns(), 1.0), X)
YD := Matrix.DotMultiplication(Y, S)
return YD
}
func CreateANN(Inputs int, NeuronsByLayer []int, Act func(*Matrix.Matrix) *Matrix.Matrix, Derivate func(*Matrix.Matrix) *Matrix.Matrix, Cost func(*Matrix.Matrix, *Matrix.Matrix) *Matrix.Matrix, DCost func(*Matrix.Matrix, *Matrix.Matrix) *Matrix.Matrix, path string) ANN {
var out ANN
out.Weights = make([]*Matrix.Matrix, len(NeuronsByLayer), len(NeuronsByLayer))
out.BestWeightsFound = make([]*Matrix.Matrix, len(NeuronsByLayer), len(NeuronsByLayer))
out.LearningRates = make([]*Matrix.Matrix, len(NeuronsByLayer), len(NeuronsByLayer))
out.Δ = make([]*Matrix.Matrix, len(NeuronsByLayer), len(NeuronsByLayer))
out.Δ1 = make([]*Matrix.Matrix, len(NeuronsByLayer), len(NeuronsByLayer))
out.ð = make([]*Matrix.Matrix, len(NeuronsByLayer)+1, len(NeuronsByLayer)+1)
out.Inputs = Inputs
out.Outputs = NeuronsByLayer[len(NeuronsByLayer)-1]
out.ActivationLayer = Act
out.DarivateActivationLayer = Derivate
out.CostFunction = Cost
out.DerviateCostFunction = DCost
out.PathWeightsInCSV = path
m := Inputs
for i := 0; i < (len(NeuronsByLayer)); i++ {
n := NeuronsByLayer[i]
// one row extra for Bias weights, we need to change to random values for this matrixes
//temp := Matrix.RandomRealMatrix(m+1, n)
out.Weights[i] = Matrix.RandomRealMatrix(m+1, n, 1.2)
out.BestWeightsFound[i] = Matrix.NullMatrixP(m+1, n)
out.LearningRates[i] = Matrix.FixValueMatrix(m+1, n, 0.0001)
//tempdelta := Matrix.NullMatrix(m+1, n)
out.ð[i] = Matrix.NullMatrix(m+1, n)
out.Δ[i] = Matrix.NullMatrixP(m+1, n)
out.Δ1[i] = Matrix.NullMatrixP(m+1, n)
m = n
}
out.AcumatedError = Matrix.NullMatrixP(m, 1)
out.MinimumErrorFound = Matrix.NullMatrixP(m, 1)
out.AcumatedError1 = Matrix.NullMatrixP(m, 1)
return out
}
