Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
15 views

140203Spring_LinearAlgebra_solution

The document outlines the midterm exam for Linear Algebra at Shiraz University, scheduled for May 2, 2024, with a duration of 90 minutes. It includes various mathematical problems related to matrices, zero-mean vectors, the rank of matrices, and differences between algorithms. The exam is structured to assess students' understanding of linear algebra concepts and their applications.

Uploaded by

yalda.hajilito
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
15 views

140203Spring_LinearAlgebra_solution

The document outlines the midterm exam for Linear Algebra at Shiraz University, scheduled for May 2, 2024, with a duration of 90 minutes. It includes various mathematical problems related to matrices, zero-mean vectors, the rank of matrices, and differences between algorithms. The exam is structured to assess students' understanding of linear algebra concepts and their applications.

Uploaded by

yalda.hajilito
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

In the name of Goodness

Midterm Exam of Linear Algebra

Spring 2024 (2nd semester of 1402-03)

Computer Sci. & Eng. & I.T. dept., Elec. & Comp. Eng. School, Shiraz University

Exam date: 2-May-24 (13/2/1403) Exam Time: 14:00 Exam Duration: 90 min.

1- Consider four matrices 𝐴ℳ×𝒩 , 𝐵𝒵×𝒲 , 𝐶𝒵×𝒩 , 𝐷ℳ×𝒲 .


a. Make a block matrix 𝐸 by them.
𝐴𝐷 𝐷𝐴 𝐶𝐵 𝐵𝐶
𝐸=[ ] 𝑜𝑟 [ ] 𝑜𝑟 [ ] 𝑜𝑟 [ ]
𝐶𝐵 𝐵𝐶 𝐴𝐷 𝐷𝐴
b. How many possible forms 𝐸 can have? Four solutions as shown above
c. What is its size? (ℳ + 𝒵) × (𝒩 + 𝒲)
d. Consider 𝐸 as an adjacency matrix of a directed graph 𝐺. What is the relation between
ℳ, 𝒩, 𝒵 & 𝒲? ℳ + 𝒵 = 𝒩 + 𝒲
e. By adding a new vertex to 𝐺 with no edge to previous vertices whereas all of them have
connected to this new vertex, what is the form of new adjacency matrix as a block matrix
of 𝐴, 𝐵, 𝐶, 𝐷 & … ?
𝐴 𝐷 1 ⃗ℳ
[𝐶 𝐵 ⃗1𝒵 ]
⃗0𝒩 ⃗0𝒲 0
2- Given an 𝒩-vector 𝑥 , try to make it zero-mean step by step.
a. Finaly compute a matrix 𝐻𝒩×𝒩 such that 𝐻𝑥 is the zero-mean form of 𝑥 .
⃗1𝑇 𝑥
𝑥̅ =
𝒩
⃗ 𝑇𝑥
1 ⃗1
1 ⃗𝑇 𝟏𝒩×𝒩
𝑥0 = 𝑥 − 1 ⃗ 𝑥̅ = 𝑥 − 1⃗ ( )=𝑥− 𝑥 = (𝐼𝒩×𝒩 − ) 𝑥 = 𝐻𝒩×𝒩 𝑥
𝒩 𝒩 𝒩
b. Compute the null-space of 𝐻.
𝑛⃗0 = 𝐻𝑛⃗ = 0⃗ ⇒ 𝑛⃗ − 1 ⃗ 𝑛̅ = 0⃗ ⇒ 𝑛⃗ = 1⃗ 𝑛̅ 𝑎𝑛𝑦 𝑣𝑒𝑐𝑡𝑜𝑟 𝑤𝑖𝑡ℎ 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑒𝑙𝑒𝑚𝑒𝑛𝑡𝑠 𝑛̅
A vector space along 1 ⃗.
c. What is the rank of 𝐻?Why?
The rank of null space (nullity) is one so the rank of 𝐻 is 𝒩 − 1
d. Prove or show that 𝐻 is idempotent (i.e. 𝐻 × 𝐻 = 𝐻)
∀𝑥 : 𝐻 × 𝐻 × 𝑥 = 𝐻 × 𝑥 0 = 𝑥 0 = 𝐻 × 𝑥
⇒ ∀𝑥 : 𝐻𝐻𝑥 = 𝐻𝑥 ⇒ 𝐻𝐻 = 𝐻
𝑜𝑟
𝟏𝒩×𝒩 𝟏𝒩×𝒩 2(𝟏𝒩×𝒩 ) (1⃗ ⃗1𝑇 )(1
⃗ ⃗1𝑇 )
𝐻𝐻 = (𝐼 − ) (𝐼 − )=𝐼− +
𝒩 𝒩 𝒩 𝒩2
2(𝟏𝒩×𝒩 ) ⃗1(1
⃗ 𝑇 ⃗1)1⃗𝑇 2(𝟏𝒩×𝒩 ) ⃗1(𝒩)1⃗𝑇 2(𝟏𝒩×𝒩 ) 𝟏𝒩×𝒩
=𝐼− + 2
= 𝐼 − + 2
=𝐼− +
𝒩 𝒩 𝒩 𝒩 𝒩 𝒩
𝟏𝒩×𝒩
=𝐼− =𝐻
𝒩

3- Assume 𝐴 & 𝐵 are outer product of 𝑎 by 𝑎 and 𝑏⃗ by 𝑏⃗, respectively (𝑒. 𝑔. 𝐴 = 𝑎𝑎𝑇 ). Prove that
square of inner product of 𝑎 & 𝑏⃗ is equal to 𝑡𝑟(𝐴𝐵).
2
𝑡𝑟(𝑎𝑎𝑇 𝑏⃗𝑏⃗ 𝑇 ) = 𝑡𝑟 ((𝑏⃗ 𝑇 𝑎)(𝑎𝑇 𝑏⃗)) = (𝑏⃗ 𝑇 𝑎)(𝑎𝑇 𝑏⃗) = (𝑎𝑇 𝑏⃗)
4- Describe at least four differences between Gauss-Elimination and Gram-Schmidt algorithms. For
example: form of independent vectors, need to pivoting, need to scaling residue, capability of
parallel computing the coefficients.
Gauss-Elimination Gram-Schmidt
Form of independence Echelon Orthonormal
Pivoting May need No
Residue scaling (normalization) No Yes
Parallel coefficient computing No Yes

Good Luck
M. Taheri

You might also like