func parseVector(strs []string) (res v.F64, err error) {
res = v.Zeroes(len(strs))
for i, str := range strs {
parsedFloat, err := strconv.ParseFloat(str, 64)
if err != nil {
return res, err
}
res[i] = parsedFloat
}
return
}
func BenchmarkLeastSquare(b *testing.B) {
f := LeastSquares([]vector.F64{
vector.F64{1, 1},
vector.F64{2, 2},
})
term := NumIterationsCrit{NumIterations: 1}
for i := 0; i < b.N; i++ {
Minimize(f, vector.Zeroes(2), 1e-8, &term, nil)
}
}
func TrainLogisticRegressionClassifier(
features []v.F64,
labels v.B,
lambda float64,
termCrit sgrad.TermCrit,
eps float64) BinaryClassifier {
y, x := sgrad.Minimize(
logisticRegressionCostFunction(features, labels, lambda),
v.Zeroes(len(features[0])),
eps,
termCrit,
nil)
return &logisticRegressionClassifier{cost: y, theta: x}
}
func TestLeastSquaresPrecise(t *testing.T) {
f := LeastSquares([]vector.F64{
vector.F64{1, 1},
vector.F64{2, 2},
})
term := RelativeMeanImprovementCrit{NumItersToAvg: 10}
v, coords := Minimize(f, vector.Zeroes(2), 1e-10, &term, nil)
t.Log("Value: ", v, "Coords: ", coords)
if math.Abs(coords[0]) > 1e-2 {
t.Error("coords[0] != 0: %s", coords[0])
}
if math.Abs(coords[1]-1) > 1e-2 {
t.Error("coords[1] != 1: %s", coords[0])
}
}
func TestLeastSquares(t *testing.T) {
// Demo problem from https://en.wikipedia.org/wiki/Linear_least_squares_(mathematics)
f := LeastSquares([]vector.F64{
vector.F64{1, 6},
vector.F64{2, 5},
vector.F64{3, 7},
vector.F64{4, 10},
})
term := RelativeMeanImprovementCrit{NumItersToAvg: 10}
v, coords := Minimize(f, vector.Zeroes(2), 1e-8, &term, nil)
t.Log("Value: ", v, "Coords: ", coords)
if math.Abs(coords[0]-3.5) > 1e-1 {
t.Error("coords[0] != 3.5: %s", coords[0])
}
if math.Abs(coords[1]-1.4) > 1e-1 {
t.Error("coords[1] != 1.4: %s", coords[0])
}
}
