Method sigmoid_train

svm.go:1518–1634

(l int, dec_values, labels, probAB []float64)

}

func (this *SVM) sigmoid_train(l int, dec_values, labels, probAB []float64) {
	var A, B float64
	prior1 := float64(0)
	prior0 := float64(0)
	var i int

	for i = 0; i < l; i++ {
		if labels[i] > 0 {
			prior1 += 1
		} else {
			prior0 += 1
		}
	}

	max_iter := 100            // Maximal number of iterations
	min_step := float64(1e-10) // Minimal step taken in line search
	sigma := float64(1e-12)    // For numerically strict PD of Hessian
	eps := float64(1e-5)
	hiTarget := float64((prior1 + 1.0) / (prior1 + 2.0))
	loTarget := float64(1 / (prior0 + 2.0))
	t := make([]float64, l)
	var fApB, p, q, h11, h22, h21, g1, g2, det, dA, dB, gd, stepsize float64
	var newA, newB, newf, d1, d2 float64
	var iter int

	// Initial Point and Initial Fun Value
	A = 0.0
	B = math.Log((prior0 + 1.0) / (prior1 + 1.0))
	fval := float64(0.0)

	for i = 0; i < l; i++ {
		if labels[i] > 0 {
			t[i] = hiTarget
		} else {
			t[i] = loTarget
		}
		fApB = dec_values[i]*A + B
		if fApB >= 0 {
			fval += t[i]*fApB + math.Log(1+math.Exp(-fApB))
		} else {
			fval += (t[i]-1)*fApB + math.Log(1+math.Exp(fApB))
		}
	}
	for iter = 0; iter < max_iter; iter++ {
		// Update Gradient and Hessian (use H' = H + sigma I)
		h11 = sigma // numerically ensures strict PD
		h22 = sigma
		h21 = 0.0
		g1 = 0.0
		g2 = 0.0
		for i = 0; i < l; i++ {
			fApB = dec_values[i]*A + B
			if fApB >= 0 {
				p = math.Exp(-fApB) / (1.0 + math.Exp(-fApB))
				q = 1.0 / (1.0 + math.Exp(-fApB))
			} else {
				p = 1.0 / (1.0 + math.Exp(fApB))
				q = math.Exp(fApB) / (1.0 + math.Exp(fApB))