Sparse Matrices

Shalini Singh Jaspal, Lecturer, BVICAM

Concept
Considerations while designing algorithms:
An application needs to choose an algorithm for solving a
problem
The choice is based upon the fact that what needs to be saved in
the target environment
Time; or
Space
Time-space complexity says that in order to save one an
algorithm, generally, needs to compromise on other; an example
storage of sparse-matrices,
Shalini Singh Jaspal, Lecturer, BVICAM

Sparse Matrices
Matrices that are sparsely (<=30%) populated.
i.e. most of whose elements are blank.
Storing these matrices in regular form implies most of the space is
wasted.
Solution: Use an alternate storage scheme in order to save space.
Cost Paid: Time complexity for accessing elements randomly
increases.

Shalini Singh Jaspal, Lecturer, BVICAM

Regular Storage Scheme

5
6

Cells Occupied: 5
Available: 5*5= 25
% Used= 5/25*100
= 20%
% Wasted= 80%
Significant as the
size of the matrix
grows

70

30

Shalini Singh Jaspal, Lecturer, BVICAM

Compressed Storage I
Row Count
Column Count

1
Actual Row Number
Actual Col Number

5
Count of
occupied cells

70

30

Shalini Singh Jaspal, Lecturer, BVICAM

Data Stored

Space Saved
Space used in original storage scheme= R*C
Space used in 3-Column notation= (N+1)*3
If N=30% of R*C =>
Space used = 90% (Approximately)
Space Saved = 10%

Shalini Singh Jaspal, Lecturer, BVICAM

Accessing Elements Sequentially

For I Between 0 To N- 1
For J Between 0 To N- 1
Get Arr[I][J]
End For
End For

Shalini Singh Jaspal, Lecturer, BVICAM

Accessing Elements Sequentially

RowCount=
ColCount=
Ctr=0
For I Between 0 To RowCount
For J Between 0 to ColCount
If <element found in Sparse> Then
Result [I][J]= Get Sparse [----][----]
Ctr++
Break <when>
Else
Result [I][J]= 0
End If
End For
End For

Shalini Singh Jaspal, Lecturer, BVICAM

Accessing Elements Randomly

Count= <Count of populated items>

For I Between 1 To Count
If <Required Row- Col is the same as Row- Col of I th
element> Then
Result= Sparse[---][---]
Break <when>
End If
End For

Shalini Singh Jaspal, Lecturer, BVICAM

Compressed Storage II
Row Count

5
0

Column Count
Problem!!!

Logical Position

5
5

10

70

23

30

Data Stored

Shalini Singh Jaspal, Lecturer, BVICAM

10

Compressed Storage II
Row Count

5
0

Column Count

Logical Position
End of Storage
Marker

5
5

10

70

23

30

-1

-1

Data Stored

Shalini Singh Jaspal, Lecturer, BVICAM

11

Space Saved
Space used in original storage scheme= R*C
Space used in 2-Column notation= (N+2)*2
If N=30% of R*C =>
Space used = 60% (Approximately)
Space Saved = 40%

Shalini Singh Jaspal, Lecturer, BVICAM

12

Accessing Elements Sequentially

Do it Yourself

Shalini Singh Jaspal, Lecturer, BVICAM

13

Accessing Elements Randomly

Do it Yourself

Shalini Singh Jaspal, Lecturer, BVICAM

14

Further Discussions
Adding two sparse matrices
Generating Transpose of a Sparse Matrix
Multiplying two Sparse Matrices

Shalini Singh Jaspal, Lecturer, BVICAM

15

This document was created with Win2PDF available at http://www.daneprairie.com.

The unregistered version of Win2PDF is for evaluation or non-commercial use only.