Skip to main content

Symmetric Positive Definite Matrix

 A real n\times n matrix A is symmetric positive definite if it is symmetric (A is equal to its transpose, A^T) and

x^T\!Ax > 0 \quad \mbox{for all nonzero vectors}~x.

By making particular choices of x in this definition we can derive the inequalities

\begin{alignedat}{2} a_{ii} &>0 \quad&&\mbox{for all}~i,\\ a_{ij} &< \sqrt{ a_{ii} a_{jj} } \quad&&\mbox{for all}~i\ne j. \end{alignedat}

Satisfying these inequalities is not sufficient for positive definiteness. For example, the matrix

A = \begin{bmatrix} 1 & 3/4 & 0 \\ 3/4 & 1 & 3/4 \\ 0 & 3/4 & 1 \\ \end{bmatrix}

satisfies all the inequalities but x^T\!Ax < 0 for x = [1,{}-\!\sqrt{2},~1]^T.

A sufficient condition for a symmetric matrix to be positive definite is that it has positive diagonal elements and is diagonally dominant, that is, a_{ii} > \sum_{j \ne i} |a_{ij}| for all i.

The definition requires the positivity of the quadratic form x^T\!Ax. Sometimes this condition can be confirmed from the definition of A. For example, if A = B^T\!B and B has linearly independent columns then x^T\!Ax = (Bx)^T Bx > 0 for x\ne 0. Generally, though, this condition is not easy to check.

Two equivalent conditions to A being symmetric positive definite are

  • every leading principal minor \det(A_k), where the submatrix A_k = A(1\colon k, 1 \colon k) comprises the intersection of rows and columns 1 to k, is positive,
  • the eigenvalues of A are all positive.

The first condition implies, in particular, that \det(A) > 0, which also follows from the second condition since the determinant is the product of the eigenvalues.

Here are some other important properties of symmetric positive definite matrices.

  • A^{-1} is positive definite.
  • A has a unique symmetric positive definite square root X, where a square root is a matrix X such that X^2 = A.
  • A has a unique Cholesky factorization A = R^T\!R, where R is upper triangular with positive diagonal elements.

Sources of positive definite matrices include statistics, since nonsingular correlation matrices and covariance matrices are symmetric positive definite, and finite element and finite difference discretizations of differential equations.

Examples of symmetric positive definite matrices, of which we display only the 4\times 4 instances, are the Hilbert matrix

H_4 = \left[\begin{array}{@{\mskip 5mu}c*{3}{@{\mskip 15mu} c}@{\mskip 5mu}} 1 & \frac{1}{2} & \frac{1}{3} & \frac{1}{4} \\[6pt] \frac{1}{2} & \frac{1}{3} & \frac{1}{4} & \frac{1}{5}\\[6pt] \frac{1}{3} & \frac{1}{4} & \frac{1}{5} & \frac{1}{6}\\[6pt] \frac{1}{4} & \frac{1}{5} & \frac{1}{6} & \frac{1}{7}\\[6pt] \end{array}\right],

the Pascal matrix

P_4 = \left[\begin{array}{@{\mskip 5mu}c*{3}{@{\mskip 15mu} r}@{\mskip 5mu}} 1 & 1 & 1 & 1\\ 1 & 2 & 3 & 4\\ 1 & 3 & 6 & 10\\ 1 & 4 & 10 & 20 \end{array}\right],

and minus the second difference matrix, which is the tridiagonal matrix

S_4 = \left[\begin{array}{@{\mskip 5mu}c*{3}{@{\mskip 15mu} r}@{\mskip 5mu}} 2 & -1 & & \\ -1 & 2 & -1 & \\ & -1 & 2 & -1 \\ & & -1 & 2 \end{array}\right].

All three of these matrices have the property that a_{ij} is non-decreasing along the diagonals. However, if A is positive definite then so is P^TAP for any permutation matrix P, so any symmetric reordering of the row or columns is possible without changing the definiteness.

A 4\times 4 symmetric positive definite matrix that was often used as a test matrix in the early days of digital computing is the Wilson matrix

W = \begin{bmatrix} 5 &7& 6& 5 \\ 7 &10& 8& 7 \\ 6& 8& 10& 9 \\ 5& 7& 9 &10 \end{bmatrix}.

What is the best way to test numerically whether a symmetric matrix is positive definite? Computing the eigenvalues and checking their positivity is reliable, but slow. The fastest method is to attempt to compute a Cholesky factorization and declare the matrix positivite definite if the factorization succeeds. This is a reliable test even in floating-point arithmetic. If the matrix is not positive definite the factorization typically breaks down in the early stages so and gives a quick negative answer.

Symmetric block matrices

C = \begin{bmatrix} A & X\\ X^T & B \end{bmatrix}

often appear in applications. If A is nonsingular then we can write

\begin{bmatrix} A & X\\ X^T & B \end{bmatrix} = \begin{bmatrix} I & 0\\ X^TA^{-1} & I \end{bmatrix} \begin{bmatrix} A & 0\\ 0 & B-X^TA^{-1}X \end{bmatrix} \begin{bmatrix} I & A^{-1}X\\ 0 & I \end{bmatrix},

which shows that C is congruent to a block diagonal matrix, which is positive definite when its diagonal blocks are. It follows that C is positive definite if and only if both A and B - X^TA^{-1}X are positive definite. The matrix B - X^TA^{-1}X is called the Schur complement of A in C.

We mention two determinantal inequalities. If the block matrix C above is positive definite then \det(C) \le \det(A) \det(B) (Fischer’s inequality). Applying this inequality recursively gives Hadamard’s inequality for a symmetric positive definite A:

\det(A) \le a_{11}a_{22} \dots a_{nn},

with equality if and only if A is diagonal.

Finally, we note that if x^TAx \ge 0 for all x\ne0, so that the quadratic form is allowed to be zero, then the symmetric matrix A is called symmetric positive semidefinite. Some, but not all, of the properties above generalize in a natural way. An important difference is that semidefinitness is equivalent to all principal minors, of which there are 2^{n-1}, being nonnegative; it is not enough to check the n leading principal minors. Consider, as an example, the matrix

\begin{bmatrix} 1 & 1 & 1\\ 1 & 1 & 1\\ 1 & 1 & 0 \end{bmatrix},

which has leading principal minors 1, 0, and 0 and a negative eigenvalue.

A complex n\times n matrix A is Hermitian positive definite if it is Hermitian (A is equal to its conjugate transpose, A^*) and x^*Ax > 0 for all nonzero vectors x. Everything we have said above generalizes to the complex case.


import numpy as np
A=np.array([[1,1,2],[2,2,3],[1,3,1]])
B=A.dot(A.T)
print("Symmetric positive definite Matrix B")
print(B)
x=np.array([1,2,3])
print("x.T A x >0")
print(x.T.dot(B.dot(x)))
eigva,eigv=np.linalg.eig(B)
print("eigen values")
print(eigva)
print("eigen vectors")
print(eigv)

o/p
Symmetric positive definite Matrix B
[[ 6 10  6]
 [10 17 11]
 [ 6 11 11]]
x.T A x >0
381
eigen values
[30.82496889  0.04141054  3.13362056]
eigen vectors
[[-0.42376703 -0.82534407 -0.37313358]
 [-0.73151695  0.55478298 -0.39635691]
 [-0.53413898 -0.10499055  0.83885191]]

Comments

Post a Comment

Popular posts from this blog

Mathematics for Machine Learning- CST 284 - KTU Minor Notes - Dr Binu V P

  Introduction About Me Syllabus Course Outcomes and Model Question Paper University Question Papers and Evaluation Scheme -Mathematics for Machine learning CST 284 KTU Overview of Machine Learning What is Machine Learning (video) Learn the Seven Steps in Machine Learning (video) Linear Algebra in Machine Learning Module I- Linear Algebra 1.Geometry of Linear Equations (video-Gilbert Strang) 2.Elimination with Matrices (video-Gilbert Strang) 3.Solving System of equations using Gauss Elimination Method 4.Row Echelon form and Reduced Row Echelon Form -Python Code 5.Solving system of equations Python code 6. Practice problems Gauss Elimination ( contact) 7.Finding Inverse using Gauss Jordan Elimination  (video) 8.Finding Inverse using Gauss Jordan Elimination-Python code Vectors in Machine Learning- Basics 9.Vector spaces and sub spaces 10.Linear Independence 11.Linear Independence, Basis and Dimension (video) 12.Generating set basis and span 13.Rank of a Matrix 14.Linear Mapping...
Post a Comment
Read more

4.3 Sum Rule, Product Rule, and Bayes’ Theorem

 We think of probability theory as an extension to logical reasoning Probabilistic modeling  provides a principled foundation for designing machine learning methods. Once we have defined probability distributions corresponding to the uncertainties of the data and our problem, it turns out that there are only two fundamental rules, the sum rule and the product rule. Let $p(x,y)$ is the joint distribution of the two random variables $x, y$. The distributions $p(x)$ and $p(y)$ are the corresponding marginal distributions, and $p(y |x)$ is the conditional distribution of $y$ given $x$. Sum Rule The addition rule states the probability of two events is the sum of the probability that either will happen minus the probability that both will happen. The addition rule is: $P(A∪B)=P(A)+P(B)−P(A∩B)$ Suppose $A$ and $B$ are disjoint, their intersection is empty. Then the probability of their intersection is zero. In symbols:  $P(A∩B)=0$  The addition law then simplifies to: $P(...
Post a Comment
Read more

5.1 Optimization using Gradient Descent

Since machine learning algorithms are implemented on a computer, the mathematical formulations are expressed as numerical optimization methods.Training a machine learning model often boils down to finding a good set of parameters. The notion of “good” is determined by the objective function or the probabilistic model. Given an objective function, finding the best value is done using optimization algorithms. There are two main branches of continuous optimization constrained and unconstrained. By convention, most objective functions in machine learning are intended to be minimized, that is, the best value is the minimum value. Intuitively finding the best value is like finding the valleys of the objective function, and the gradients point us uphill. The idea is to move downhill (opposite to the gradient) and hope to find the deepest point. For unconstrained optimization, this is the only concept we need,but there are several design choices. For constrained optimization, we need to intr...
Post a Comment
Read more