Artificial Neural Networks – Apply-Level Questions & Model Answers
Set 1 – Activation Functions
Question (15 Marks): Q1. Apply the ReLU and Sigmoid activation functions to a simple two-layer
neural network. Show how the output differs for an input vector X = [1, -2, 3]. Interpret the result.
Model Answer / Solution:
Solution:
- ReLU Output ≈ 4.6
- Sigmoid Output ≈ 1.084
- ReLU passes positives unchanged, zeroes negatives → larger linear response.
- Sigmoid squashes inputs to (0,1) → smaller, dampened output.
- ReLU speeds convergence, Sigmoid risks vanishing gradients.
Marking Scheme (Rubric):
Criteria Marks
Stepwise application of functions 6
Correct outputs 5
Interpretation 4
Set 2 – Neural Networks
Question (15 Marks): Q2. Given a classification problem with handwritten digit images, apply the
appropriate neural network type (Feedforward, CNN, or RNN). Justify your choice and outline how
you would structure the layers.
Model Answer / Solution:
Solution:
- Appropriate choice: CNN.
- Justification: Exploits spatial locality, translation invariance.
- Layer structure example: Input 28x28 → Conv+ReLU → Pool → Conv+ReLU → Pool → Flatten →
Dense+ReLU → Dense+Softmax.
- Training: Cross-entropy loss, Adam optimizer, augmentation for invariance.
Marking Scheme (Rubric):
Criteria Marks
Correct choice of NN 5
Layer structure outline 6
Justification/application 4
Set 3 – Cardinality
Question (15 Marks): Q3. Apply the concept of cardinality of fuzzy relations to R={(a,b,0.7),
(a,c,0.4), (b,c,0.9)}. Calculate the cardinality and explain its implication.
Model Answer / Solution:
Solution:
- Fuzzy cardinality = 0.7+0.4+0.9 = 2.0
- Normalized cardinality = 2.0/4 = 0.5
- Average membership = 2.0/3 ≈ 0.667
- Interpretation: moderate overall strength, covers 50% of relation space.
Marking Scheme (Rubric):
Criteria Marks
Cardinality calculation 6
Normalization/averaging 4
Interpretation 5
Set 4 – Fuzzy Relation Operations
Question (15 Marks): Q4. Consider R={(x,y,0.6),(x,z,0.8)} and S={(y,w,0.7),(z,w,0.9)}. Apply the
max–min composition to find R■S. Show steps.
Model Answer / Solution:
Solution:
- Path x→y→w: min(0.6,0.7)=0.6
- Path x→z→w: min(0.8,0.9)=0.8
- Max=0.8
- Result: R■S={(x,w,0.8)}
Marking Scheme (Rubric):
Criteria Marks
Correct application of max–min 7
Stepwise calculation 5
Final result 3
Set 5 – Fuzzy Relation Properties
Question (15 Marks): Q5. For R={(a,a,1),(a,b,0.7),(b,a,0.7),(b,b,1)}, check whether it is symmetric
and/or transitive (max–min).
Model Answer / Solution:
Solution:
- Symmetry: Yes, since µ(a,b)=µ(b,a).
- Transitivity: Check via R■R ≤ R. Holds for all pairs.
- Conclusion: Symmetric and transitive; fuzzy equivalence relation.
Marking Scheme (Rubric):
Criteria Marks
Symmetry check 5
Transitivity check 7
Conclusion 3
