AI Question Bank · Series 1 — Solved
Mathematical and Statistical Foundation of Data Science
AI-generated practice bank for PGD01C01 — 100 Part A (2 marks) + 20 Part B (20 marks) questions covering all five modules, fully solved in a simple exam-oriented style. Click any question to reveal the answer.
How to use this bank
Click a question to reveal its answer. Each question matches the exam style of the Calicut PGDDSA paper — short crisp answers for 2-mark questions, full statements + derivations + worked examples for 20-mark questions.
Coverage. 20 Part-A + 4 Part-B per module × 5 modules.
| Module | Topic |
|---|---|
| 1 | Set theory, relations, combinatorics |
| 2 | Graph theory |
| 3 | Linear algebra |
| 4 | Vectors & geometry of space |
| 5 | Probability theory |
PART A — Short Answer (2 marks each) · 100 questions
Module 1 — Set Theory, Relations, Combinatorics
A set is a well-defined collection of distinct objects called elements. Example: — the set of positive integers less than 5.
The power set is the set of all subsets of . If , then . For : .
. We add both and subtract the overlap counted twice.
A binary relation that is reflexive (), symmetric (), and transitive ().
A partition of is a collection of non-empty, pairwise-disjoint subsets whose union is . Every equivalence relation creates a partition.
A set in bijection with (countably infinite). Examples: . Note: is uncountable.
For any set , . No set is in bijection with its own power set; hence is uncountable.
A relation that is reflexive, antisymmetric () and transitive. Example: divisibility on .
.
.
Letters: S×3, T×3, A×1, I×2, C×1 (total 10). Arrangements .
Circular arrangements . Fix one person and arrange the rest.
. .
and . The complement of union = intersection of complements.
— set of all ordered pairs. .
A relation on is reflexive if for every . Example: on (every number is itself).
for all . Example: "is a sibling of" — if A is sibling of B, then B is sibling of A.
and . Example: "is taller than" — if A is taller than B and B taller than C, then A is taller than C.
If (i) is true and (ii) for all , then is true for all natural numbers .
such subsets.
— the number of ways to arrange all distinct objects.
(empty selection) and (the full set). Both equal 1.
. Choosing from = choosing that include a fixed element + that exclude it.
. Picking to include = picking to exclude.
. (By Binomial theorem with .) This equals the total number of subsets of an -element set.
11 letters: E×3, N×3, G×2, I×2, R×1. Arrangements .
Arrange 5 boys: . This creates 6 gaps (before, between, after). Choose 4 of 6 gaps for the girls and arrange them: . Total .
Pick 4 of 5 digits and arrange: four-digit numbers.
Each of the 4 positions has 26 choices independently: words.
ways.
Two specific books are pre-selected. Choose remaining 2 from the other 8: ways.
Necklace arrangements (rotations + reflections equivalent): arrangements.
"EQUATION" has 8 letters, 5 vowels (E, U, A, I, O) and 3 consonants (Q, T, N). Treat the 5 vowels as one block: 4 units to arrange = . Vowels inside the block: . Total .
Stars and bars: ways.
Total committees of 5 from 14: . Subtract committees with fewer than 2 women — i.e., 0 women or 1 woman:
- 0 women: .
- 1 woman: .
Answer ways.
Module 2 — Graph Theory
A graph with no self-loops and no multiple edges between the same pair of vertices.
A graph that allows multiple edges between the same pair of vertices, but typically no self-loops.
. Each edge contributes 1 to the degree of each endpoint, so the sum is twice the edge count.
— every pair of vertices joined.
— every vertex on the left joined to every vertex on the right.
A closed walk that uses every edge exactly once. A connected graph has one iff every vertex has even degree (Euler's theorem).
A cycle that visits every vertex of the graph exactly once. Finding one is NP-complete in general.
A path is a sequence of distinct vertices joined by edges; first and last differ. A cycle is a closed path (first vertex = last vertex).
, where counts all faces including the unbounded outer face.
For simple planar graphs . For : , but . So cannot be planar.
A subgraph of a connected graph that is a tree and includes every vertex of . It has edges.
Exactly . A tree is connected and acyclic, with this minimum number of edges to connect all vertices.
For vertices, the matrix where if and 0 otherwise. Symmetric for undirected graphs.
The number of walks of length exactly from vertex to vertex .
A graph where each edge has a direction (ordered pair). Each vertex has separate in-degree and out-degree.
In-degree of a vertex = number of edges coming into it. Out-degree = number of edges going out from it.
A graph is connected if there is a path between every pair of vertices. Otherwise, it splits into connected components.
A graph whose vertex set splits into two disjoint subsets such that every edge connects to . No odd cycles.
A vertex whose removal disconnects the graph (or increases the number of components).
An edge whose removal disconnects the graph (or increases components). Removing it breaks a connected graph.
Module 3 — Linear Algebra
The rank is the number of linearly independent rows (equivalently, columns). It equals the dimension of the column space.
For an matrix : .
Diagonal matrix — eigenvalues are the diagonal entries: .
A non-zero vector such that for some scalar (the eigenvalue). only stretches — doesn't rotate it.
Trace = sum of diagonal entries: . Equals the sum of all eigenvalues.
A square matrix with — it has no inverse, its columns are linearly dependent, and 0 is an eigenvalue.
Any real matrix can be written , where and are orthogonal and is diagonal with non-negative singular values.
A square matrix with — its columns (and rows) form an orthonormal set. .
— measures the angle between two vectors. Used in text and embedding similarity.
The set for some non-zero . It's an -dimensional flat. In 2D it's a line; in 3D a plane.
— the set of vectors mapped to zero. Its dimension is the nullity.
Vectors are linearly independent if implies all .
Every square matrix satisfies its own characteristic equation: if , then .
.
A square matrix with 1's on the diagonal and 0's elsewhere; acts as identity for matrix multiplication: .
For square with , the inverse satisfies . For : .
A square matrix with , i.e., . Always has real eigenvalues and orthogonal eigenvectors.
A symmetric matrix with for all non-zero . All eigenvalues positive.
— the inner product is bounded by the product of norms.
Dropping a perpendicular from a vector onto a subspace. For vector onto : .
Module 4 — Vectors and Geometry of Space
The vector from origin to a point . For , .
.
.
. Magnitude = = area of parallelogram.
for . Direction = .
, where is the normal and is a point on the plane.
For plane and point : .
For line through with direction and point : .
— sphere of radius centred at .
, , .
, , .
. Take for the angle.
.
— a scalar equal to the (signed) volume of the parallelepiped with edges .
— absolute value of the scalar triple product.
Find two edge vectors from one point, compute their cross product (gives normal ), then use .
, where is a point on the line and its direction.
.
. Unit vector .
. Direction cosines: .
Module 5 — Probability Theory
(i) . (ii) . (iii) For disjoint events, .
for . "Probability of given has occurred."
. Inverts conditioning using prior and likelihood.
and are independent iff . Knowing doesn't change .
A function taking countably many values, described by PMF with .
A function taking values in an interval, described by PDF with . Probabilities = .
. Mean = , variance = .
for .
Mean . Variance .
. Average the conditional means over .
. Measures linear co-movement.
Mean = variance = .
for , else 0. Mean = , variance = .
for . Mean = , variance = .
For any random variable with finite variance: .
. Non-decreasing, right-continuous, .
Mean = , variance = . The Poisson is the only standard distribution where mean = variance.
Mean = . Variance = .
If partitions , then .
. Measures the strength of linear association.
PART B — Long Essay (20 marks each) · 20 questions
Module 1 — Set Theory, Relations, Combinatorics
Statement. For three finite sets :
Proof. Take any element , suppose it lies in exactly of the three sets.
- LHS counts once.
- RHS counts : singletons − pair-intersections + triple.
For : net = ; ; . So each element counted once on both sides. ∎
Application. Let = multiples of in , so .
- .
- .
- .
- .
- .
.
So 271 integers in 1–500 are divisible by 3, 5 or 7.
Definition. A relation on that is reflexive, symmetric, and transitive. For , the equivalence class .
Theorem. The equivalence classes of form a partition of .
Proof.
- Non-empty. Reflexivity: , so .
- Cover. Each , so .
- Disjoint or equal. Suppose , pick in both. Then and . Symmetry gives , transitivity . For any : , so . So . By symmetry . ∎
Verification with on , .
- Reflexive: , divisible by 4. ✓
- Symmetric: . ✓
- Transitive: and . ✓
The four classes are — residue classes mod 4 — and they partition .
Principle. If is true (base case) and (induction step), then is true for all .
Part 1. Prove .
Base : LHS = 1, RHS = . ✓
Step. Assume . Then . ∎
Part 2. Prove for .
Base : , divisible by 3. ✓
Step. Assume , i.e., for some integer . Then — divisible by 3. ∎
(i) Choose 4 of 10 men and 3 of 8 women: ways.
(ii) "MATHEMATICS" has 11 letters: M×2, A×2, T×2, H, E, I, C, S. Arrangements .
(iii) Treat the 2 specific persons as a single block: 5 units to arrange = . The 2 within the block can swap: . Total ways.
Module 2 — Graph Theory
Theorem. A connected graph has an Eulerian circuit iff every vertex has even degree.
Proof. (⇒) If an Eulerian circuit exists, every time the walk enters a vertex it must leave by another edge. Edges at each vertex pair up, so every degree is even.
(⇐) Suppose connected and all even degree. Start a trail at any vertex . The even-degree condition ensures that whenever we reach a vertex, an unused outgoing edge exists. The trail must return to , forming a circuit .
If uses all edges, done. Otherwise, by connectivity, some vertex on has unused edges; build another circuit and splice in. Repeat until all edges used. ∎
Verify . : every vertex has degree (odd!). So does NOT have an Eulerian circuit. (Only with odd does — degree even.)
Hamiltonian. Visit vertices in cyclic order . Every consecutive pair is joined (complete graph). So DOES have a Hamiltonian circuit. Every with is Hamiltonian.
Algorithm.
Initialise d[v] = ∞ for all v; d[source] = 0
Use min-priority queue keyed by d[v]
while queue not empty:
u = extract-min
for each neighbour v of u:
if d[u] + w(u,v) < d[v]:
d[v] = d[u] + w(u,v)
prev[v] = u
Greedy choice: always finalise unvisited vertex with smallest tentative distance. Complexity with a heap.
Apply.
| Step | Visited | d[A] | d[B] | d[C] | d[D] | d[E] |
|---|---|---|---|---|---|---|
| init | – | 0 | ∞ | ∞ | ∞ | ∞ |
| pick A | {A} | 0 | 6 | 2 | ∞ | ∞ |
| pick C | {A,C} | 0 | 5* | 2 | 6** | 10 |
| pick B | {A,B,C} | 0 | 5 | 2 | 6 | 10 |
| pick D | {A,B,C,D} | 0 | 5 | 2 | 6 | 9*** |
| pick E | all | 0 | 5 | 2 | 6 | 9 |
*via (cost ). **via (cost ). ***via (cost ).
Shortest path: , total cost 9.
Definitions.
- Spanning tree: subgraph that is a tree and includes all vertices; has edges.
- MST: spanning tree with minimum total edge weight.
Kruskal's algorithm.
sort edges by weight ascending
Union-Find: each vertex its own component initially
for each edge (u, v) in order:
if find(u) != find(v):
add (u, v) to MST
union(u, v)
stop when |V| - 1 edges added
Apply. Sort edges: .
| # | Edge | Weight | Action |
|---|---|---|---|
| 1 | (B, D) | 1 | add — {B,D} |
| 2 | (A, C) | 2 | add — {A,C}, {B,D} |
| 3 | (B, C) | 3 | add — merges to {A,B,C,D} |
| 4 | (D, E) | 3 | add — {A,B,C,D,E} |
| stop | 4 edges = $ | V | - 1$ |
MST edges: . Total weight = .
Definitions.
- Adjacency matrix : if , else 0. Symmetric for undirected graphs.
- Incidence matrix : if vertex is on edge .
Key fact. = number of walks of length exactly from to .
Apply.
Compute . Row 1:
- .
- .
- .
- .
By symmetry of , is symmetric. Continuing:
Interpretation. From vertex 1: 2 walks back to 1 (via 2 or 3), 1 walk to 2 (via 3), 1 walk to 3 (via 2), 2 walks to 4 (via 2 or 3). Diagonal entries equal vertex degrees.
Module 3 — Linear Algebra
Step 1 — Characteristic polynomial. .
Eigenvalues: . Check: sum , product . ✓
Step 2 — Eigenvectors.
For : . Take .
For : . Take .
Step 3 — Diagonalise. .
, so .
Step 4 — Compute . where .
.
Process. For :
- Form .
- Find eigenvalues and orthonormal eigenvectors of → right singular vectors .
- Singular values , sorted descending → .
- Left singular vectors → .
- Then .
Apply.
Step 1. .
Step 2. .
Eigenvectors:
- : .
- : .
Step 3. .
Step 4. Left singular vectors: .
.
Result. .
Theorem. For an matrix : .
Proof. Let = nullity, a basis of . Extend to basis of .
Claim: is a basis of the column space.
- Spans: any has , so (first killed).
- Independent: if then , so this lies in . But it's in the complement basis, so all .
Hence rank = , giving rank + nullity = . ∎
Apply. Row-reduce:
One non-zero row ⇒ rank = 1. By rank-nullity: nullity = 3 - 1 = 2.
Null space spanned by and (or similar two-vector basis).
Inner product. For : . Satisfies bilinearity, symmetry, positivity. Induces norm and angle .
Cosine similarity = — scale-invariant. Used for text similarity, embeddings, recommender systems.
Distance metrics. Take , .
| Metric | Formula | Value | Use |
|---|---|---|---|
| Euclidean () | k-means, k-NN, generic | ||
| Manhattan () | $\sum | x_i - y_i | $ |
| Cosine | Text, embeddings, direction matters | ||
| Mahalanobis | depends on | Correlated features, anomaly detection |
Summary. Euclidean is the default; Manhattan when outliers; Cosine for direction; Mahalanobis when features are correlated/have different scales.
Module 4 — Vectors and Geometry of Space
Dot product. . With basis where , expansion gives .
Cross product. is perpendicular to both, magnitude , direction by right-hand rule. Using etc: .
Computations.
Dot: .
Cross: .
Magnitudes: , .
Angle: . .
Parallelogram area: .
Derivation. For plane with normal , and point . Let = foot of perpendicular from to . Then , so . Substituting into the plane:
.
Distance = . ∎
Plane through .
Edge vectors: , .
Normal : .
Plane equation using : (after multiplying by ).
Distance from origin. .
Derivation. For skew lines and .
The common perpendicular direction is . The shortest distance equals the absolute projection of onto :
Apply. , , .
.
.
.
.
Scalar triple product. .
Geometric meaning. Base = parallelogram with edges , area , and perpendicular to it. The height equals the projection of onto . Hence
Volume = base × height = .
Equivalently as a determinant: .
If this determinant is zero, the three vectors are coplanar.
Apply. .
Volume = cubic units.
Since the determinant is non-zero, the three vectors are not coplanar — they truly span a 3D parallelepiped.
Module 5 — Probability Theory
Bayes' theorem. If partitions with and has :
Proof. From definition: . So . By law of total probability: . Substituting gives Bayes' formula. ∎
Disease problem.
Let = "has disease", = "tests positive". Given:
- .
- Sensitivity: .
- Specificity: , so false-positive .
Total probability: .
Bayes: .
So only about 8.76% chance the patient actually has the disease! Even a highly accurate test fails when the disease is rare — false positives dominate. This is the base-rate fallacy.
Types of random variables.
A random variable assigns a real number to each outcome. Two main types:
Discrete RV. Takes countably many values. Described by PMF with .
Continuous RV. Takes values in an interval. Described by PDF with .
Distributions.
| Distribution | PMF / PDF | Mean | Variance | Use |
|---|---|---|---|---|
| Bernoulli | Single yes/no trial | |||
| Binomial | # successes in trials | |||
| Poisson | Rare event counts | |||
| Uniform | on | Equal-likelihood | ||
| Exponential | Waiting times | |||
| Normal | Natural variation, errors |
Importance. Bernoulli is the basic unit; Binomial = sum of Bernoullis. Poisson approximates Binomial when large, small with . The Normal arises naturally by the Central Limit Theorem — sums of many independent random variables tend to Normal regardless of original distribution.
Algorithm. Given target density , easy-to-sample proposal , and constant with for all :
loop:
X ~ g
U ~ Uniform(0, 1)
if U <= f(X) / (M * g(X)): accept X, return
else: reject, repeat
Proof. Let = "accept this iteration".
.
So and — accepted has the target CDF. ∎
Application. Target on . Normalisation: , so for .
Proposal: (uniform on ).
is maximised at : . So works.
Algorithm:
- .
- .
- Accept if .
Acceptance rate — very efficient.
Variance identity. . ∎
Linearity of expectation. For any random variables and scalars : .
This holds without any independence assumption. For variance, requires .
Binomial mean and variance via linearity.
Write , where each independent.
Mean. . By linearity: .
Variance. . By independence: .
For .
- Mean = .
- Variance = .
- Standard deviation = .
So a Binomial has expected value 8 and typical fluctuations of about around 8.
End of question bank. Total: 100 Part-A (200 marks possible) + 20 Part-B (400 marks possible) = 600 marks of practice material. All five modules covered with 20 + 4 questions each.