2025 USA-NA-AIO Round 2, Problem 3, Part 2

AI Olympiads 2025 USA-NA-AIO Round 2
1

Part 2 (5 points, coding task)

This dataset is too big. In our contest, we only use a small portion with 1000 samples.

To avoid introducing any bias, we will randomly select 1000 distinct samples.

Use NumPy to randomly select 1000 sample indices.

  • Use the random seed number 2025 to generated randomized indices. After the generation is completed, reset the seed number back to None.

  • The name of the output is called indices. It must be a list that contains 1000 integer type (not numpy array integers) objects.

2 
### WRITE YOUR SOLUTION HERE ###

np.random.seed(2025)
indices = np.random.permutation(len(dataset_train))[:1000]
np.random.seed()
indices = [int(idx) for idx in indices]

""" END OF THIS PART """