Checklist USA-NA-AIO 2025 Round 1 · Task 2
Neural Network Basics
Work through affine transformations, their gradients and NumPy/PyTorch linear modules, then train an MLP on a harmonic-motion regression set.
The task
Problem 2 (100 points, 13 parts) is described as 'about the basics of neural network'. It covers a numeric affine transformation, gradients with respect to x, W and b, a NumPy module My_Linear_NumPy handling inputs of arbitrary leading shape, a two-layer linear PyTorch model, a symmetric tied-weight network and the rank of W^T W, a custom ReLU module, an MLP class, and finally a synthetic harmonic-motion dataset of 1,000 samples on which the MLP is trained with MSE and Adam using full-batch training.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You submit
- Written answers and code in a notebook.
- Rules
- Only numpy, matplotlib and torch (as imported in the starter cell).
- All coding tasks run on CPU.
- Format
- 2025 USA-NA-AIO Round 1, 24 March 2025 (date from IOAI news article). Statements and official solutions were posted part by part on the USAAIO forum on 28 March 2025.