Government Engineering College, Khagaria

Mid-Semester Examination
Subject: Artificial Intelligence (155503)
Semester: 5th Max Marks: 20 Time: 2 Hours
Instructions:
1. There are 20 marks in total. It’s like a treasure hunt, but the treasure is knowledge!
2. Be creative. Show us your problem-solving skills (and not just Google’s)!
3. Write legibly or use your best handwriting magic. We need to read your answers, not guess them.
4. If you can’t figure out a question, please don’t panic. We’re all in this together. (But still try!)

Note: AI can predict your marks... if you show your working properly. Don’t try to confuse the
machine (or the examiner).

Section A: Objective Questions (1 mark 7) DFID is a search technique that combines:

each) A. DFS and BFS B. DFS and A* C. BFS and UCS
D. Hill climbing and DFS
Attempt any five out of seven. Each carries 1
mark. Use your brain — ChatGPT isn’t allowed in
Section B: Situation-Based Questions (2
exams (yet).
marks each)
1) Which of the following is an unsupervised Attempt any two out of three. Real-life chaos
learning algorithm? meets machine learning — can you sort it out like
A. Decision Tree B. K-Means C. Linear a proper AI guru? Justify your answer, or risk
Regression D. Neural Network being replaced by a chatbot.

2) What is the primary difference between DFS 8) Garbage Gurus Unite!

and BFS? The city is drowning in trash. Literally. Garbage
A. Memory usage B. Accuracy C. Type of nodes trucks are now equipped with GPS and weight
D. Cost function sensors — fancy! But alas, the data is as
unlabeled as your Tupperware. Which magical
3) Which of the following is not a component of technique will help officials know where to drop
an artificial neuron? the bins like it's hot?
A. Weights B. Bias C. Activation Function D. A. Decision Tree B. K-Means C. Supervised
Cluster Index Learning
4) Semi-supervised learning is best defined as: 9) The Dropout Dilemma
A. Only labeled data used B. Only unlabeled An online learning platform is panicking.
data used C. Uses both labeled and unlabeled Students log in, watch half a video, then
data D. Data learned by feedback disappear like your morning motivation. Some
are labeled as 'likely to drop out.' How do we find
5) In Decision Trees, information gain is used to:
the rest of these academic ninjas before it’s too
A. Prune the tree B. Select the best feature for
late?
split C. Update neuron weights D. Normalize
A. Supervised Learning B. Semi-supervised
input data
Learning C. Unsupervised Learning
6) What does an epoch refer to in deep learning?
10) Rise of the Warehouse Wanderer
A. One full pass through training data B. A
A robot is sent to find items in a massive
hidden layer C. Number of outputs D. Error
warehouse. It doesn’t know the map, the shelves
correction formula
move (somehow), and the robot only wants to
reach snacks... I mean, parts. Which search path Section E: Feedforward Neural Network (4
should this digital explorer take? marks)
A. BFS B. DFS C. DFID
15) The Great Neural Saga
Just like your semester, this network goes in one
Section C: Define/Differentiate (2 marks
direction — forward. No backprop, no regrets.
each)
Attempt any two. Be clear, concise, and confident You are given:
— as if explaining to your confused-but-curious Input Layer: 2 neurons → X₁ = 1, X₂ = 0 (the only
examiner. time 0 effort gives you a pass!)
Hidden Layer: 2 neurons
11) Differentiate between Supervised and
- Neuron 1: Weights = [0.5, -0.4], Bias = 0.1
Unsupervised Learning with examples. Bonus
- Neuron 2: Weights = [0.3, 0.8], Bias = -0.2
points if your example includes a dog.
Output Layer: 1 neuron
12) Define a Decision Tree and explain its - Weights from hidden neurons = [0.6, -0.1], Bias
working with an example. Drawing trees outside 0.05
=
the exam hall doesn’t count. - Activation Function: Sigmoid → σ(x) = 1 / (1 +
e^(-x))
13) Differentiate between Breadth First Search
(BFS) and Depth-First Iterative Deepening Calculate the final output of this network. Show
(DFID). DFS’s responsible cousin. every step, every neuron, every heartbeat and
diagram (obviously!).
Section D: Numerical Problem (3 marks)
Attempt any one. Show your work. Calculators
are allowed, crystal balls are not.

14) (a) Apply K-means clustering to the 1D data

points: [2, 4, 10, 12, 20] using k=2 and initial
centroids as 2 and 20. Show one complete
iteration with updated centroids. No shortcuts —
we see you.
OR

(b) The weather is unpredictable, but your AI

model shouldn’t be. Use the data below to build
the root node of a Decision Tree. Calculate
information gain and pick the feature that splits
like a Bollywood plot.

| Weather | Temperature | Play |

|---------|-------------|------|
| Sunny | Hot | No |
| Overcast| Cool | Yes |
| Rainy | Mild | Yes |
| Sunny | Cool | No |
| Rainy | Mild | Yes |