Let A_{mXn} be a given matrix with m > n. Find upper bound on the number of additions and multiplications required to determine the rank using the elimination procedure.

Basic Understanding of Problem Nature

Consider matrix

    \[ \left A_{5X3} = \right \left (\begin{array}{rrr} a_{11} & a_{12} & a_{13} \\ a_{21} & a_{22} & a_{23} \\ a_{31} & a_{32} & a_{33} \\ a_{41} & a_{42} & a_{43} \\ a_{51} & a_{52} & a_{53} \end{array} \right) \]

This simulates A_{mXn} matrix with m > n.

Now, let’s consider Gauss Elimination. Every elementary row operation is of the form R_2 \leftarrow R_2 + (m_{21})*R_1.

(1)   \begin{equation*} \begin{minipage}{1\textwidth} So there is one addition & one multiplication per elementary row operation. Also, number of additions=multiplications. \end{minipage} \end{equation*}

Gauss Elimination for matrix A:

Step 1:

a_{11} is pivot for 1^{st} row. So, a_{21}, a_{31}, a_{41} and a_{51} should be made zero. For this, we need to perform elementary row operations on 5-1 = 4 rows. Each row will have 3+1-1=3 elements.

Step 2:

a_{22} is pivot for 2^{nd} row. So, a_{32}, a_{42} and a_{52} should be made zero. For this, we need to perform elementary row operations on 5-2 = 3 rows. Each row will have 3+1-2 = 2 elements. This is because the 1st column elements of all rows except 1 are now zero. This excluded 3 elements.

Step 3:

a_{33} is pivot for 3^{rd} row. So, a_{43} and a_{53} should be made zero. For this, we need to perform elementary row operations on 5-3 = 2 rows. Each row will have 3+1-3 = 1 element. This is because the 1^{st} & 2^{nd} column elements of all rows except a_{11} & a_{22} are now zero. This excluded 4 elements.

(2)   \begin{equation*} \begin{minipage}{1\textwidth} Thus for the kth step, when $a_{kk}$ is the pivot, the elementary row operations will be done on $m-k$ rows and each row will have $n+1-k$ elements. \end{minipage} \end{equation*}

Solution

From (1) & (2),

Total no. of operations (add + mult) for Gauss Elimination,

    \begin{align*} \left T(n) = 2 * \sum_{k=1}^{n}{1(m-k)(n-k+1)} \end{align*}

(3)   \begin{equation*} \left = 2 * \sum_{k=1}^{n}{k^{2} - k(m+n+1) + (mn+m)} \end{equation*}

Let c = m + n + 1 and d = mn + m. Apply this in (3),

    \begin{align*} \left T(n) = 2 * \sum_{k=1}^{n}{k^{2} - ck + d} \end{align*}

Applying rules of summation & ignoring lower order variables,

    \begin{align*} \left T(n) = [2.n(n+1)(2n+1)] / 6 \end{align*}

    \begin{align*} \left = (2n^3 + 3n^2 + n) / 3 \end{align*}

Therefore, Upper bound on the number of additions and multiplications required to determine the rank using the elimination procedure,

    \begin{align*} \left T(n) \leq O(2n^3/3) \end{align*}

Degree of highest order polynomial is 3.

Subscribe to Ehan Ghalib!

Invalid email address
We promise not to spam you. You can unsubscribe at any time.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes:

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>