add shifted eigenvalue procedure and unit test

This commit is contained in:
Elizabeth Hunt 2023-11-27 14:06:16 -07:00
parent 76e00682b2
commit 793c01c9bd
Signed by: simponic
GPG Key ID: 52B3774857EB24B1
5 changed files with 128 additions and 66 deletions

View File

@ -1035,22 +1035,22 @@ double dominant_eigenvalue(Matrix_double *m, Array_double *v, double tolerance,
return lambda; return lambda;
} }
#+END_SRC #+END_SRC
*** ~least_dominant_eigenvalue~ *** ~shift_inverse_power_eigenvalue~
+ Author: Elizabeth Hunt + Author: Elizabeth Hunt
+ Name: ~least_dominant_eigenvalue~ + Name: ~least_dominant_eigenvalue~
+ Location: ~src/eigen.c~ + Location: ~src/eigen.c~
+ Input: a pointer to an invertible matrix ~m~, an initial eigenvector guess ~v~ (that is non + Input: a pointer to an invertible matrix ~m~, an initial eigenvector guess ~v~ (that is non
zero or orthogonal to an eigenvector with the dominant eigenvalue), a ~tolerance~ and zero or orthogonal to an eigenvector with the dominant eigenvalue), a ~shift~ to act as the
~max_iterations~ that act as stop conditions shifted \delta, and ~tolerance~ and ~max_iterations~ that act as stop conditions.
+ Output: the least dominant eigenvalue with the lowest magnitude, approximated with the Inverse + Output: the eigenvalue closest to ~shift~ with the lowest magnitude closest to 0, approximated
Power Iteration Method with the Inverse Power Iteration Method
#+BEGIN_SRC c #+BEGIN_SRC c
double least_dominant_eigenvalue(Matrix_double *m, Array_double *v, double shift_inverse_power_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, size_t max_iterations) { double shift, double tolerance,
size_t max_iterations) {
assert(m->rows == m->cols); assert(m->rows == m->cols);
assert(m->rows == v->size); assert(m->rows == v->size);
double shift = 0.0;
Matrix_double *m_c = copy_matrix(m); Matrix_double *m_c = copy_matrix(m);
for (size_t y = 0; y < m_c->rows; ++y) for (size_t y = 0; y < m_c->rows; ++y)
m_c->data[y]->data[y] = m_c->data[y]->data[y] - shift; m_c->data[y]->data[y] = m_c->data[y]->data[y] - shift;
@ -1065,22 +1065,38 @@ double least_dominant_eigenvalue(Matrix_double *m, Array_double *v,
Array_double *normalized_eigenvector_2 = Array_double *normalized_eigenvector_2 =
scale_v(eigenvector_2, 1.0 / linf_norm(eigenvector_2)); scale_v(eigenvector_2, 1.0 / linf_norm(eigenvector_2));
free_vector(eigenvector_2); free_vector(eigenvector_2);
eigenvector_2 = normalized_eigenvector_2;
Array_double *mx = m_dot_v(m, eigenvector_2); Array_double *mx = m_dot_v(m, normalized_eigenvector_2);
double new_lambda = double new_lambda =
v_dot_v(mx, eigenvector_2) / v_dot_v(eigenvector_2, eigenvector_2); v_dot_v(mx, normalized_eigenvector_2) /
v_dot_v(normalized_eigenvector_2, normalized_eigenvector_2);
error = fabs(new_lambda - lambda); error = fabs(new_lambda - lambda);
lambda = new_lambda; lambda = new_lambda;
free_vector(eigenvector_1); free_vector(eigenvector_1);
eigenvector_1 = eigenvector_2; eigenvector_1 = normalized_eigenvector_2;
} }
return lambda; return lambda;
} }
#+END_SRC #+END_SRC
*** ~least_dominant_eigenvalue~
+ Author: Elizabeth Hunt
+ Name: ~least_dominant_eigenvalue~
+ Location: ~src/eigen.c~
+ Input: a pointer to an invertible matrix ~m~, an initial eigenvector guess ~v~ (that is non
zero or orthogonal to an eigenvector with the dominant eigenvalue), a ~tolerance~ and
~max_iterations~ that act as stop conditions.
+ Output: the least dominant eigenvalue with the lowest magnitude closest to 0, approximated
with the Inverse Power Iteration Method.
#+BEGIN_SRC c
double least_dominant_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, size_t max_iterations) {
return shift_inverse_power_eigenvalue(m, v, 0.0, tolerance, max_iterations);
}
#+END_SRC
*** ~leslie_matrix~ *** ~leslie_matrix~
+ Author: Elizabeth Hunt + Author: Elizabeth Hunt
+ Name: ~leslie_matrix~ + Name: ~leslie_matrix~

View File

@ -21,7 +21,23 @@ finds the least dominant eigenvalue on the matrix:
0 & 2 & 6 0 & 2 & 6
\end{bmatrix} \end{bmatrix}
which has eigenvalues: $5 + \sqrt{17}, 2, 5 - \sqrt{17}$ and should produce $\sqrt{17}$. which has eigenvalues: $5 + \sqrt{17}, 2, 5 - \sqrt{17}$ and should thus produce $5 - \sqrt{17}$.
See also the entry ~Eigen-Adjacent -> least_dominant_eigenvalue~ in the LIZFCM API See also the entry ~Eigen-Adjacent -> least_dominant_eigenvalue~ in the LIZFCM API
documentation. documentation.
* Question Four
See ~UTEST(eigen, shifted_eigenvalue)~ in ~test/eigen.t.c~ which
finds the least dominant eigenvalue on the matrix:
\begin{bmatrix}
2 & 2 & 4 \\
1 & 4 & 7 \\
0 & 2 & 6
\end{bmatrix}
which has eigenvalues: $5 + \sqrt{17}, 2, 5 - \sqrt{17}$ and should thus produce $2.0$.
With the initial guess: $[0.5, 1.0, 0.75]$.
See also the entry ~Eigen-Adjacent -> shift_inverse_power_eigenvalue~ in the LIZFCM API
documentation.

View File

@ -76,6 +76,9 @@ extern double fixed_point_secant_bisection_method(double (*f)(double),
extern double dominant_eigenvalue(Matrix_double *m, Array_double *v, extern double dominant_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, size_t max_iterations); double tolerance, size_t max_iterations);
extern double shift_inverse_power_eigenvalue(Matrix_double *m, Array_double *v,
double shift, double tolerance,
size_t max_iterations);
extern double least_dominant_eigenvalue(Matrix_double *m, Array_double *v, extern double least_dominant_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, double tolerance,
size_t max_iterations); size_t max_iterations);

View File

@ -49,12 +49,12 @@ double dominant_eigenvalue(Matrix_double *m, Array_double *v, double tolerance,
return lambda; return lambda;
} }
double least_dominant_eigenvalue(Matrix_double *m, Array_double *v, double shift_inverse_power_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, size_t max_iterations) { double shift, double tolerance,
size_t max_iterations) {
assert(m->rows == m->cols); assert(m->rows == m->cols);
assert(m->rows == v->size); assert(m->rows == v->size);
double shift = 0.0;
Matrix_double *m_c = copy_matrix(m); Matrix_double *m_c = copy_matrix(m);
for (size_t y = 0; y < m_c->rows; ++y) for (size_t y = 0; y < m_c->rows; ++y)
m_c->data[y]->data[y] = m_c->data[y]->data[y] - shift; m_c->data[y]->data[y] = m_c->data[y]->data[y] - shift;
@ -69,17 +69,22 @@ double least_dominant_eigenvalue(Matrix_double *m, Array_double *v,
Array_double *normalized_eigenvector_2 = Array_double *normalized_eigenvector_2 =
scale_v(eigenvector_2, 1.0 / linf_norm(eigenvector_2)); scale_v(eigenvector_2, 1.0 / linf_norm(eigenvector_2));
free_vector(eigenvector_2); free_vector(eigenvector_2);
eigenvector_2 = normalized_eigenvector_2;
Array_double *mx = m_dot_v(m, eigenvector_2); Array_double *mx = m_dot_v(m, normalized_eigenvector_2);
double new_lambda = double new_lambda =
v_dot_v(mx, eigenvector_2) / v_dot_v(eigenvector_2, eigenvector_2); v_dot_v(mx, normalized_eigenvector_2) /
v_dot_v(normalized_eigenvector_2, normalized_eigenvector_2);
error = fabs(new_lambda - lambda); error = fabs(new_lambda - lambda);
lambda = new_lambda; lambda = new_lambda;
free_vector(eigenvector_1); free_vector(eigenvector_1);
eigenvector_1 = eigenvector_2; eigenvector_1 = normalized_eigenvector_2;
} }
return lambda; return lambda;
} }
double least_dominant_eigenvalue(Matrix_double *m, Array_double *v,
double tolerance, size_t max_iterations) {
return shift_inverse_power_eigenvalue(m, v, 0.0, tolerance, max_iterations);
}

View File

@ -1,50 +1,7 @@
#include "lizfcm.test.h" #include "lizfcm.test.h"
UTEST(eigen, leslie_matrix) { Matrix_double *eigen_test_matrix() {
Array_double *felicity = InitArray(double, {0.0, 1.5, 0.8}); // produces a matrix that has eigenvalues [5 + sqrt{17}, 2, 5 - sqrt{17}]
Array_double *survivor_ratios = InitArray(double, {0.8, 0.55});
Matrix_double *m = InitMatrixWithSize(double, 3, 3, 0.0);
m->data[0]->data[0] = 0.0;
m->data[0]->data[1] = 1.5;
m->data[0]->data[2] = 0.8;
m->data[1]->data[0] = 0.8;
m->data[2]->data[1] = 0.55;
Matrix_double *leslie = leslie_matrix(survivor_ratios, felicity);
EXPECT_TRUE(matrix_equal(leslie, m));
free_matrix(leslie);
free_matrix(m);
free_vector(felicity);
free_vector(survivor_ratios);
}
UTEST(eigen, leslie_matrix_dominant_eigenvalue) {
Array_double *felicity = InitArray(double, {0.0, 1.5, 0.8});
Array_double *survivor_ratios = InitArray(double, {0.8, 0.55});
Matrix_double *leslie = leslie_matrix(survivor_ratios, felicity);
Array_double *v_guess = InitArrayWithSize(double, 3, 2.0);
double tolerance = 0.0001;
uint64_t max_iterations = 64;
double expect_dominant_eigenvalue = 1.22005;
double approx_dominant_eigenvalue =
dominant_eigenvalue(leslie, v_guess, tolerance, max_iterations);
EXPECT_NEAR(expect_dominant_eigenvalue, approx_dominant_eigenvalue,
tolerance);
free_vector(v_guess);
free_vector(survivor_ratios);
free_vector(felicity);
free_matrix(leslie);
}
UTEST(eigen, least_dominant_eigenvalue) {
Matrix_double *m = InitMatrixWithSize(double, 3, 3, 0.0); Matrix_double *m = InitMatrixWithSize(double, 3, 3, 0.0);
m->data[0]->data[0] = 2.0; m->data[0]->data[0] = 2.0;
m->data[0]->data[1] = 2.0; m->data[0]->data[1] = 2.0;
@ -54,12 +11,17 @@ UTEST(eigen, least_dominant_eigenvalue) {
m->data[1]->data[2] = 7.0; m->data[1]->data[2] = 7.0;
m->data[2]->data[1] = 2.0; m->data[2]->data[1] = 2.0;
m->data[2]->data[2] = 6.0; m->data[2]->data[2] = 6.0;
return m;
}
UTEST(eigen, least_dominant_eigenvalue) {
Matrix_double *m = eigen_test_matrix();
double expected_least_dominant_eigenvalue = 0.87689; // 5 - sqrt(17) double expected_least_dominant_eigenvalue = 0.87689; // 5 - sqrt(17)
double tolerance = 0.0001; double tolerance = 0.0001;
uint64_t max_iterations = 64; uint64_t max_iterations = 64;
Array_double *v_guess = InitArrayWithSize(double, 3, 2.0); Array_double *v_guess = InitArrayWithSize(double, 3, 1.0);
double approx_least_dominant_eigenvalue = double approx_least_dominant_eigenvalue =
least_dominant_eigenvalue(m, v_guess, tolerance, max_iterations); least_dominant_eigenvalue(m, v_guess, tolerance, max_iterations);
@ -88,3 +50,63 @@ UTEST(eigen, dominant_eigenvalue) {
free_matrix(m); free_matrix(m);
free_vector(v_guess); free_vector(v_guess);
} }
UTEST(eigen, shifted_eigenvalue) {
Matrix_double *m = eigen_test_matrix();
double least_dominant_eigenvalue = 0.87689; // 5 - sqrt{17}
double dominant_eigenvalue = 9.12311; // 5 + sqrt{17}
double expected_middle_eigenvalue = 2.0;
double shift = (dominant_eigenvalue + least_dominant_eigenvalue) / 2.0;
double tolerance = 0.0001;
uint64_t max_iterations = 64;
Array_double *v_guess = InitArray(double, {0.5, 1.0, 0.75});
double approx_middle_eigenvalue = shift_inverse_power_eigenvalue(
m, v_guess, shift, tolerance, max_iterations);
EXPECT_NEAR(approx_middle_eigenvalue, expected_middle_eigenvalue, tolerance);
}
UTEST(eigen, leslie_matrix_dominant_eigenvalue) {
Array_double *felicity = InitArray(double, {0.0, 1.5, 0.8});
Array_double *survivor_ratios = InitArray(double, {0.8, 0.55});
Matrix_double *leslie = leslie_matrix(survivor_ratios, felicity);
Array_double *v_guess = InitArrayWithSize(double, 3, 2.0);
double tolerance = 0.0001;
uint64_t max_iterations = 64;
double expect_dominant_eigenvalue = 1.22005;
double approx_dominant_eigenvalue =
dominant_eigenvalue(leslie, v_guess, tolerance, max_iterations);
EXPECT_NEAR(expect_dominant_eigenvalue, approx_dominant_eigenvalue,
tolerance);
free_vector(v_guess);
free_vector(survivor_ratios);
free_vector(felicity);
free_matrix(leslie);
}
UTEST(eigen, leslie_matrix) {
Array_double *felicity = InitArray(double, {0.0, 1.5, 0.8});
Array_double *survivor_ratios = InitArray(double, {0.8, 0.55});
Matrix_double *m = InitMatrixWithSize(double, 3, 3, 0.0);
m->data[0]->data[0] = 0.0;
m->data[0]->data[1] = 1.5;
m->data[0]->data[2] = 0.8;
m->data[1]->data[0] = 0.8;
m->data[2]->data[1] = 0.55;
Matrix_double *leslie = leslie_matrix(survivor_ratios, felicity);
EXPECT_TRUE(matrix_equal(leslie, m));
free_matrix(leslie);
free_matrix(m);
free_vector(felicity);
free_vector(survivor_ratios);
}