forward.go

package nn

import "gonum.org/v1/gonum/mat"

// forward takes input data and returns the 'activation' and 'z'
// from each layer - z = w.x +b and a = sigmoid(z)
func (n *MLP) forward(x mat.Matrix) (as, zs []mat.Matrix) {
    as = append(as, x) // first activation is input
    
    _x := x
    
    for i := 0; i < len(n.weights); i++ {
        w := n.weights[i]
        b := n.biases[i]
        
        
        dot := new(mat.Dense)
        dot.Mul(_x, w)
        
        z := new(mat.Dense)
        addB := func(_, col int, v float64) float64 { return v + b.At(col, 0) }
        z.Apply(addB, dot)
        
        a := new(mat.Dense)
        a.Apply(applySigmoid, z)
        
        zs = append(zs, z)
        as = append(as, a)
        
        _x = a
    }
  
    return
}