Indian Institute of Technology Madras
BS Degree Programme · Examination & Study Paper
Mathematical Foundations of Generative AI
BiGAN, domain adaptation & FID
Official Paper
Examination Instructions & Guidelines
- Solve on notebook/paper first before checking solutions.
- MCQ: Exactly 1 option is correct. Negative marking applies.
- MSQ: One or more options correct. Select all to score.
- Scientific calculator allowed for all calculations.
- Official answer key is sealed at the end of this sheet.
Quiz 1 · 26 Oct 2025 (AN)Quiz 1
- Q1MCQ2 Marks·Choose one option

- A−0.8008
- B−0.6432
- C−1.1412
- D−0.4794
- Q2MCQ2 Marks·Choose one option

- A0.22
- B0.28
- C0.34
- D0.41
- Q3MCQ2 Marks·Choose one option

- AThe first term measures bias shift; the second term measures covariance mismatch.
- BBoth terms measure covariance only.
- CBoth terms are dominated by the generator’s variance.
- DThe first term measures variance difference; the second measures mean shift.
- Q4MCQ2 Marks·Choose one option
FID compares real and generated image distributions by assuming they follow multivariate Gaussian distributions in a feature space extracted from a pretrained neural network. Which layer’s activations are typically used to compute this feature space in practice?
- AThe output logits of the generator
- BRaw RGB pixel values
- CThe first convolution layer of the discriminator
- DThe penultimate layer of an Inception-v3 network
- Q5MCQ2 Marks·Choose one option

- A

- B

- C

- D

- A
- Q6MCQ3 Marks·Choose one option

- A[−2.1, 0.4]
- B[−2.2, −4.3]
- C[0.2, 0.1]
- D[0, 0]
- Q7MSQ4 Marks·Choose all correct options

- A

- B

- C

- D

- A
End Term · 31 Aug 2025 (FN)End Term
- Q8MCQ4 Marks·Choose one option

- A

- B

- C

- D

- A