func ExampleCholesky() {
// Construct a symmetric positive definite matrix.
tmp := mat64.NewDense(4, 4, []float64{
2, 6, 8, -4,
1, 8, 7, -2,
2, 2, 1, 7,
8, -2, -2, 1,
})
var a mat64.SymDense
a.SymOuterK(1, tmp)
fmt.Printf("a = %0.4v\n", mat64.Formatted(&a, mat64.Prefix(" ")))
// Compute the cholesky factorization.
var chol mat64.Cholesky
if ok := chol.Factorize(&a); !ok {
fmt.Println("a matrix is not positive semi-definite.")
}
// Find the determinant.
fmt.Printf("\nThe determinant of a is %0.4g\n\n", chol.Det())
// Use the factorization to solve the system of equations a * x = b.
b := mat64.NewVector(4, []float64{1, 2, 3, 4})
var x mat64.Vector
if err := x.SolveCholeskyVec(&chol, b); err != nil {
fmt.Println("Matrix is near singular: ", err)
}
fmt.Println("Solve a * x = b")
fmt.Printf("x = %0.4v\n", mat64.Formatted(&x, mat64.Prefix(" ")))
// Extract the factorization and check that it equals the original matrix.
var t mat64.TriDense
t.LFromCholesky(&chol)
var test mat64.Dense
test.Mul(&t, t.T())
fmt.Println()
fmt.Printf("L * L^T = %0.4v\n", mat64.Formatted(&a, mat64.Prefix(" ")))
// Output:
// a = ⎡120 114 -4 -16⎤
// ⎢114 118 11 -24⎥
// ⎢ -4 11 58 17⎥
// ⎣-16 -24 17 73⎦
//
// The determinant of a is 1.543e+06
//
// Solve a * x = b
// x = ⎡ -0.239⎤
// ⎢ 0.2732⎥
// ⎢-0.04681⎥
// ⎣ 0.1031⎦
//
// L * L^T = ⎡120 114 -4 -16⎤
// ⎢114 118 11 -24⎥
// ⎢ -4 11 58 17⎥
// ⎣-16 -24 17 73⎦
}
// StdDev predicts the standard deviation of the function at x.
func (g *GP) StdDev(x []float64) float64 {
if len(x) != g.inputDim {
panic(badInputLength)
}
// nu_* = k(x_*, k_*) - k_*^T * K^-1 * k_*
n := len(g.outputs)
kstar := mat64.NewVector(n, nil)
for i := 0; i < n; i++ {
v := g.kernel.Distance(g.inputs.RawRowView(i), x)
kstar.SetVec(i, v)
}
self := g.kernel.Distance(x, x)
var tmp mat64.Vector
tmp.SolveCholeskyVec(g.cholK, kstar)
var tmp2 mat64.Vector
tmp2.MulVec(kstar.T(), &tmp)
rt, ct := tmp2.Dims()
if rt != 1 || ct != 1 {
panic("bad size")
}
return math.Sqrt(self-tmp2.At(0, 0)) * g.std
}
// ConditionNormal returns the Normal distribution that is the receiver conditioned
// on the input evidence. The returned multivariate normal has dimension
// n - len(observed), where n is the dimension of the original receiver. The updated
// mean and covariance are
// mu = mu_un + sigma_{ob,un}^T * sigma_{ob,ob}^-1 (v - mu_ob)
// sigma = sigma_{un,un} - sigma_{ob,un}^T * sigma_{ob,ob}^-1 * sigma_{ob,un}
// where mu_un and mu_ob are the original means of the unobserved and observed
// variables respectively, sigma_{un,un} is the unobserved subset of the covariance
// matrix, sigma_{ob,ob} is the observed subset of the covariance matrix, and
// sigma_{un,ob} are the cross terms. The elements of x_2 have been observed with
// values v. The dimension order is preserved during conditioning, so if the value
// of dimension 1 is observed, the returned normal represents dimensions {0, 2, ...}
// of the original Normal distribution.
//
// ConditionNormal returns {nil, false} if there is a failure during the update.
// Mathematically this is impossible, but can occur with finite precision arithmetic.
func (n *Normal) ConditionNormal(observed []int, values []float64, src *rand.Rand) (*Normal, bool) {
if len(observed) == 0 {
panic("normal: no observed value")
}
if len(observed) != len(values) {
panic("normal: input slice length mismatch")
}
for _, v := range observed {
if v < 0 || v >= n.Dim() {
panic("normal: observed value out of bounds")
}
}
ob := len(observed)
unob := n.Dim() - ob
obMap := make(map[int]struct{})
for _, v := range observed {
if _, ok := obMap[v]; ok {
panic("normal: observed dimension occurs twice")
}
obMap[v] = struct{}{}
}
if len(observed) == n.Dim() {
panic("normal: all dimensions observed")
}
unobserved := make([]int, 0, unob)
for i := 0; i < n.Dim(); i++ {
if _, ok := obMap[i]; !ok {
unobserved = append(unobserved, i)
}
}
mu1 := make([]float64, unob)
for i, v := range unobserved {
mu1[i] = n.mu[v]
}
mu2 := make([]float64, ob) // really v - mu2
for i, v := range observed {
mu2[i] = values[i] - n.mu[v]
}
n.setSigma()
var sigma11, sigma22 mat64.SymDense
sigma11.SubsetSym(n.sigma, unobserved)
sigma22.SubsetSym(n.sigma, observed)
sigma21 := mat64.NewDense(ob, unob, nil)
for i, r := range observed {
for j, c := range unobserved {
v := n.sigma.At(r, c)
sigma21.Set(i, j, v)
}
}
var chol mat64.Cholesky
ok := chol.Factorize(&sigma22)
if !ok {
return nil, ok
}
// Compute sigma_{2,1}^T * sigma_{2,2}^-1 (v - mu_2).
v := mat64.NewVector(ob, mu2)
var tmp, tmp2 mat64.Vector
err := tmp.SolveCholeskyVec(&chol, v)
if err != nil {
return nil, false
}
tmp2.MulVec(sigma21.T(), &tmp)
// Compute sigma_{2,1}^T * sigma_{2,2}^-1 * sigma_{2,1}.
// TODO(btracey): Should this be a method of SymDense?
var tmp3, tmp4 mat64.Dense
err = tmp3.SolveCholesky(&chol, sigma21)
if err != nil {
return nil, false
}
tmp4.Mul(sigma21.T(), &tmp3)
for i := range mu1 {
mu1[i] += tmp2.At(i, 0)
}
// TODO(btracey): If tmp2 can constructed with a method, then this can be
// replaced with SubSym.
for i := 0; i < len(unobserved); i++ {
for j := i; j < len(unobserved); j++ {
v := sigma11.At(i, j)
sigma11.SetSym(i, j, v-tmp4.At(i, j))
}
}
return NewNormal(mu1, &sigma11, src)
}
