You are on page 1of 6

Summary of Module 1

by P Vanchinathan

• A linear system of equations has 3 possibilities: (i) No solutions (ii)


Only one Solution (iii) ∞-ly many solutions. (So a system with exactly
2 solutions or 95 solutions etc is not possible).

• If the RHS of all the equations is zero we call it a homogeneous


system and it always has a solution: a solution where all variables are
zero. It is called the trivial solution and theoretically very convenient
and may not be useful in practice.

• Row operations are three kinds.

1. Multiply a row by non-zero number.


2. Add a row to another row
3. Interchange two rows.

• Every system of linear equations/matrix can be converted to


triangular form by suitable row operation.

• Trianguar form means: all entries below (1,1) cell, (2,2) cell, (3,3) cells
etc are zero.

• Theorem: Any system of equations and the triangular system obtained


from it by row operations have both same set of solutions.

• TIP: To convert to triangular form we should aim to make entries zero


in specific order. Start from first column, work downwards, then the
second column work downwards . . .

• For a non-homegeneous system we write the coefficients of the vari-


ables in matrix form (one rwo per equation) also the rhs as a column
(separated by a vertical line for clarity) and apply row operation. For
homogeneous system as rhs will always remain zero under all row op-
erations the rhs is omitted from calculations.

• By a solution we mean a specification of values for all the variables


which upon substitution makes all the equations valid.

1
NON-HOMOGENEOUS SYSTEM:

• If the triangular form ends with last row all zero including rhs there
are infinitely many solutions.

• If the last row is zero except the rhs then system has NO SOLUTION.

• Unique solution happens if the last row has exactly one nonzero entry in
LHS. By back substitution all varaible values can be calculated one by
one. Possible only when number of equations and number of variables
are equal.

2
SOLVING HOMOGENEOUS SYTEMS:
Unique solution or Infinitely many solutions?

• After bringing an equation to triangular form scan each row and mark
the first non-zero entry of each row (say by circling them or using a
special colour).

• Now scan each column to see if any entry in that column is marked.

• If a column contains a marked entry then it represents a basic variable


and other columns represent free variables.

• A homogeneous system has infinitely many solutions if and only there


are free variables.

• Working from the last row obtain formula for all basic variables in
terms of free variables.

• Now in a solution (in the format of a vector) each component is either


a free variable or a formula using free variables alone.

• Specify value for all free variables arbitrarily and build a solution vector.
Thus infinite number of solutions are obtained.

• Special solutions are obtained by setting one free variable value as 1


and other free variable with value 0.

• Repeat this procedure for setting each free variable value 1 in turns.

• So we obtain one solution vector each for every free variable.

• Any solution is obtained by using these solutions and forming linear


combinations;

• By a linear combination of a set of vectors we mean multiplying each


vector by some scalar and then adding them up.

3
• Notation: ei represents a column vector with 1 in its ith slot and zero
in all other slots (the size of the vector get specified according to the
context).

• For any matrix A, it is simple calculation to see that multiplication


matrix with vector ei , that is Aei gives the i-th column vector of A

• If eTi represents the transpose vector (so it will be a row vector then
eTi A will be the i-th row vector of A.

• Calling the individual column vector of a matrix A as v1 , v2 , . . . vn (that


is A = (v1 | v1 | · · · | vn ) multiplying A be a general column vector
X = (x1 , x2 , . . . , xn )T gives

AX = x1 v1 +x2 v2 +· · · xn vn , a linear combn of column vectors of A

• Interpretation of solving a system AX = b: Finding a solution means


checking if there is a linear combination of column vectors of A that
produces the rhs vector b.

4
ELEMENTARY MATRICES

• When elementary row operation is applied to the identity matrix we


get an ELEMENTARY matrix.

• Determinant of first kind of elementary matrix is the same as the con-


stant used for multiplying the row.
 
1 0 0
8
e.g., det  0 − 83 0 = −
 
3
0 0 1

• Determinant for second kind (add a row to another row) is +1.


 
1 0 0
e.g., det  1 1 0  = +1
 

0 0 1

• Determinat for third kind (interchange two rows) is always −1.

1 0 0 0
 
0 0 0 1
e.g., det   = −1

0 0 1 0

0 1 0 0

• Determinat of a trinagular matrix is simply the product of its diag-


onal entries.

• Determinant of any matrix by Row operations:

– Carry out row operation As to convert to triangular form, T .


– For every row operation used cumulatively keep multiplying the
determinats of the corresponding elementary matrices, E1 , E2 , . . .
arriving at a final answer x.
– When you stop at triangular form T calculcate its determinant of
T.
– det(A) = det(T )/x.

• Elementary matrices of first and second kind are lower triangular.

5
• Product of lower triangular matrices is again lower triangular.

• Inverse of lower triangular matrix is again lower triangular!

• Similarly for upper triangular.

• A triangular matrix is a product of Diagonal matrix and a uni-triangular


matrix.

• Gaussian elimination as matrix equation:

A = E1−1 E2−1 . . . Ek−1 DU

• If no interchange of rows was used then A = LDU (This is called


LU/LDU decomposition of a matrix)

• By a permuation matrix we mean a matrix obtained by permuting the


columns wholesale. What is AP ?

1. Once we have a factorization A = LU , this helps us to solve AX = b


for many vectors b in a much easier way.

2. To solve AX = b is same finding vector X satisfying LU X = b

3. In the beginning X = (x1 , x2 , . . . , xn )T is unknown and hence the vector


U X is unknown, and call it U X = Y = (y1 , y2 , . . . , yn )T .

4. So our original equation becomes Ly = b

5. So solve this system LY = b for Y . This is easy as L is (lower) trian-


gular.

6. After the above step now Y has become known.

7. Now solve U X = Y . This is again easy as U is upper triangular. The


answer obtained at this step will also solve original system AX = b.

You might also like