Week-8

Mathematics for Data Science - 2

Projection, Gram-Schmidt process, Orthogonal transformation
Graded Assignment-Solutions

1 Multiple Select Questions (MSQ)

1. An inner product on R3 is defined as:

⟨., .⟩ : R3 × R3 → R
⟨(x1 , x2 , x3 ), (y1 , y2 , y3 )⟩ = x1 y1 + x2 y2 + x3 y3 .

Match the sets of vectors in column A with their properties of orthogonality or orthonor-
mality in column B with respect to the above inner product.

Set of vectors Properties

(Column A) (Column B)

a) {(2, 3, 4), (−1, 2, −1)} i) Forms a basis but not orthogonal

b) { √12 (1, 0, −1), √12 (−1, 0, −1)} ii) Forms an orthogonal basis

c) {(2, 3, 4), (−1, 2, −1), (0, 4, −3)} iii) Orthogonal but not orthonormal,
and does not form a basis of R3

d) {(2, 3, 4), (−1, 2, −1), (11, 2, −7)} iv) Orthonormal,

but does not form a basis of R3

Table : M2W6G1

Choose the correct options.

 ⃝ Option 1: a → iv)
⃝ Option 2: a → iii)
⃝ Option 3: b → iv)
⃝ Option 4: b → iii)
⃝ Option 5: c → ii)
⃝ Option 6: c → i)
⃝ Option 7: d → i)
⃝ Option 8: d → ii)
Solution: Note that if x, y ∈ V are non-zero and orthogonal, then {x, y} is a linearly
independent set. Also, since dimension of R3 is 3, any set of 3 linearly independent
vectors forms a basis.

a) ⟨(2, 3, 4), (−1, 2, −1)⟩ = −2 + 6 − 4 = 0 and ⟨(2, 3, 4), (2, 3, 4)⟩ ̸= 1. so the set is
orthogonal but not orthonormal. Also, since there are only 2 vectors, it cannot
form a basis for R3 . This matches with iii).

b) ⟨ √12 (1, 0, −1), √12 (−1, 0, 1)⟩ = 0. Also ⟨ √12 (1, 0, −1), √12 (1, 0, −1)⟩ = 1 and
⟨ √12 (−1, 0, 1), √12 (−1, 0, 1)⟩ = 1. Thus these vectors are orthonormal. But this does
not form a basis for R3 , since there are only two vectors. This matches with iv).

c) ⟨(−1, 2, −1), (0, 4, −3)⟩ = 11 ̸= 0, so the set is not orthogonal. If we show that
the set is linearly independent, then it forms a basis for R3 (this is because a lin-
early independent set with number of elements equal to dim(V ) is a basis for V ).
To show that the set is linearly independent. show that the system α(2, 3, 4) +
β(−1, 2, −1) + γ(0, 4, −3) = 0 has only one solution α = β = γ = 0. This matches
with i).

d) ⟨(2, 3, 4), (−1, 2, −1)⟩ = ⟨(2, 3, 4), (11, 2, −7)⟩ = ⟨(−1, 2, −1), (11, 2, −7)⟩ = 0. So
the set is orthogonal and hence is linearly independent. Thus it forms a basis (since
the linearly independent set has 3(=dim(R3 )) elements). This matches with ii).

2
2. Choose the set of correct options.
⃝ Option 1: Suppose β = {v1 , v2 , . . . , vn } is an orthogonal basis of an inner
product space V . If there exists some v ∈ V , such that ⟨v, vi ⟩ = 0 for all
i = 1, 2, . . . , n, then v = 0.
⃝ Option 2: There exists an orthonormal basis for Rn with the standard inner
product.
⃝ Option 3: If PW denotes the linear transformation which projects the vec-
tors of an inner product space V to a subspace W of V , then range(PW ) ∩
null space(PW ) = {0}, where 0 denotes the zero vector of V .
 
1 −1
⃝ Option 4: cannot represent a matrix corresponding to some projec-
0 1
tion.
Solution: A linear transformation P is a projection if P 2 = P .

• Option 1: Since β is a basis, v ∈ V can be written as v = α1 v1 + α2 v2 + ... + αn vn .

Now ⟨v, vi ⟩ = αi ⟨vi , vi ⟩ for all i = 1, 2, ..., n. Since it is given that v is such that
⟨v, vi ⟩ = 0 for all i, we have αi ∥vi ∥2 = 0 for all i. But ∥vi ∥ ≠ 0 (since vi ̸= 0). Thus
αi = 0 for all i and hence v = 0.
• Option 2: {e1 , e2 , ..., en } is an orthonormal basis for V .
• Option 3: Since PW is a projection, PW2
= PW . Let x ∈ range(PW ) ∩ null space(PW ).
Since x ∈ range(PW ), x = PW y for some y ∈ V . But x ∈ null space(PW ) implies
2 2
that PW x = 0, i.e., PW y = 0. But PW = PW and hence x = PW y = 0. Thus
range(PW ) ∩ null space(PW ) = {0}.
 2    
1 −1 1 −2 1 −1
• Option 4: Note that = ̸= . So the matrix cannot
0 1 0 1 0 1
represent a projection.

2 Numerical Answer Type (NAT)

3. If A is an orthogonal matrix of order 5, then find nullity of the matrix A. [Ans: 0]

Solution: A is orthogonal of order 5 =⇒ A is invertible. Therefore nullity of A is 0.

4. Let v ∈ R3 be a vector such that ||v|| = 5. If u is the vector obtained from v after the
anti- clock wise rotation of XY-plane with angle 70◦ about the Z- axis, then find the
length of the vector u. [Ans: 5]

Solution: Rotation matrix is an orthogonal matrix. It preserves the length of a vector,

that is, ||Av|| = ||v|| for all v. Therefore ||v|| = ||u|| = 5.

3
5. Let v = (1, 2, 2) be a vector in R3 . If (a, b, c) is the vector obtained from v after the anti-
clock wise rotation of YZ-plane with angle 60◦ about the X- axis, then find the value of
a + b + c. [Ans: 3]
 
1 0 0
Solution: A = 0 cos(60) sin(60)  represents the rotation matrix of XY -plane with
0 sin(60) cos(60)
◦
angle 60 about the X-axis.
      
1 0 0 1 a 1√
0 cos(60) − sin(60) 2 =  b  = 1 − 3 .
√
0 sin(60) cos(60) 2 c 1+ 3
Hence a + b + c = 3.
6. Consider a vector space M2×2 (R) and a norm on the  vectorspace defined as
a a
∥A∥ = max{|a11 | + |a21 |, |a12 | + |a22 |}, where A = 11 12 ∈ M2×2 (R).
a21 a22
 √ 
x 2x
Let B =  be an orthogonal matrix i.e. BB T = B T B = I and assume
√

− 2y y
x, y > 0.
√ √ 
3x 3x
Then find the norm of the matrix C = √  (i.e., ||C||) [Ans: 2]
√
3y 3y
 √  √   2 
x 2x x − 2y 3x 0
Solution: Note that BB T =  √ 
√
 =  . But
2
− 2y y 2x y 0 3y
T 2 2
√ √
since B is an orthogonal matrix, BB = I. So 3x = 3y = 1. Thus 3x = 3y = ±1.
This gives ∥C∥ = max{2, 2} = 2.
7. Let V = R2 be the inner product space with usual inner product and a linear transfor-
mation T : R2 → R2 defined as T (x, y) = ( √aa2 +4 x + √b21+1 y, √a22 +4 x + √b2b+1 y). If T is an
orthogonal linear transformation, then find the value of a + 2b. [Ans: 0]

Solution: If T is an orthogonal transformation, then ⟨T u, T v⟩ = ⟨u, v⟩, for all u =

(u1 , u2 ) and v = (v1 , v2 ). Thus we have
⟨T u, T v⟩ = ⟨(u1 , u2 ), (v1 , v2 )⟩
a 1 a 1
(√ u1 + √ u2 )( √ v1 + √ v2 )
2
a +4 2
b +1 2
a +4 2
b +1
2 b 2 b
+ (√ u1 + √ u2 )( √ v1 + √ v2 ) = u1 v1 + u2 v2
a2 + 4 b2 + 1 a2 + 4 b2 + 1

4
 a + 2b
Thus we get p p (u1 v2 +u2 v1 ) = 0. Since ⟨T u, T v⟩ = ⟨u, v⟩ is true for all
(a2+ 4) (b2 + 1)
a + 2b
u, v, we can choose u = v = (1, 1). For this choice of u, v, we have p p =
(a2 + 4) (b2 + 1)
0. Thus a + 2b = 0.

3 Comprehension Type Question:

With a particular frame of reference (in R3 ), position of a target is given as the vector
(3, 4, 5). Three shooters S1 , S2 , and S3 are moving along the lines x = y, x = −y, and
x = 2y, on the XY -plane (i.e., z = 0) to shoot the target. Suppose that, there is another
shooter S4 , who is moving on the plane x + y + z = 0. Suppose all of them shoot the target
so that the target is at the closest distance from the respective path or plane on which they
are travelling.
Answer questions 8,9 and 10 using the given information.

8. Choose the set of correct options.

⃝ Option 1: S1 will shoot the target from the point ( 27 , − 72 , 0).
⃝ Option 2: S1 will shoot the target from the point ( 72 , 27 , 0).
⃝ Option 3: S1 will shoot the target from the point (1, 1, 0).
⃝ Option 4: S2 will shoot the target from the point (− 21 , − 12 , 0).
⃝ Option 5: S2 will shoot the target from the point (1, −1, 0).
⃝ Option 6: S2 will shoot the target from the point (− 12 , 12 , 0).
⃝ Option 7: S3 will shoot the target from the point (4, 2, 0).
⃝ Option 8: S3 will shoot the target from the point (2, 1, 0).
⃝ Option 9: S3 will shoot the target from the point (0, 0, 0).
Solution: Note that the closest point to shoot the target can be obtained by finding
the orthogonal projection of (3,4,5) onto the subspace on which Si ’s are moving.

Equation of line of motion of S1 is x = y (x − y = 0). The subspace W1 = {(x, y, z)|x =

y, z = 0} is spanned by the vector (1, 1, 0). The projection of (3, 4, 5) onto W1 is
1
⟨(1,1,0),(1,1,0)⟩
⟨(3, 4, 5), (1, 1, 0)⟩(1, 1, 0) = ( 72 , 27 , 0).

Equation of line of motion of S2 is x = −y (x+y = 0). The subspace W2 = {(x, y, z)|x =

−y, z = 0} is spanned by the vector (1, −1, 0). The projection of (3, 4, 5) onto W2 is
1
⟨(1,−1,0),(1,−1,0)⟩
⟨(3, 4, 5), (1, −1, 0)⟩(1, −1, 0) = (− 21 , 21 , 0).

5
 Equation of line of motion of S3 is x = 2y (x−2y = 0). The subspace W3 = {(x, y, z)|x =
2y, z = 0} is spanned by the vector (2, 1, 0). The projection of (3, 4, 5) onto W3 is
1
⟨(2,1,0),(2,1,0)⟩
⟨(3, 4, 5), (2, 1, 0)⟩(2, 1, 0) = (4, 2, 0).

9. If (a, b, c) is the point from which the shooter S4 will shoot the target, then find the
value of a + 2b + 3c. [Ans: 2]
Solution: Equation of line of motion of S4 is x + y + z = 0. The subspace W4 =
{(x, y, z)|x + y + z = 0} is spanned by the vectors (1, 0, −1) and (0, 1, −1). Using Gram
Schmidt orthogonalization, we can get an orthonormal (orthogonal is sufficient) basis for
W4 . {(1, 0, −1), (− 21 , 1, − 12 )} is an orthogonal basis for W4 . The projection of (3, 4, 5)
onto W4 is
1 1
⟨(1,0,−1),(1,0,−1)⟩
⟨(3, 4, 5), (1, 0, −1)⟩(1, 0, −1)+ ⟨(− 1 ,1,− 1 ),(− 1
,1,− 12 )⟩
⟨(3, 4, 5), (− 21 , 1, − 12 )⟩(− 12 , 1, − 12 )
2 2 2
= (−1, 0, 1) + (0, 0, 0) = (−1, 0, 1).

10. Let di be the distance of the target from the point where the shooter Si shoots the target,
for i = 1,2,3,4 and let d be the minimum amongst the di . Find the value of d2 . [Ans:
25.5]
p
Solution: Distanceq between the points can be calculated using (x1 − x2 )2 + (y1 − y2 )2 + (z1 − z2 )2 .
q √ √
Thus we have d1 = 51 2
, d2 = 99 2
, d3 = 30 and d4 = 4 3.