Data Preprocessing Pipeline

[maximizemaxwell]

Issue 1 — Data Preprocessing Pipeline

Description: Prepare Fashion-MNIST datasets (T2, T4, T10) with PCA reduction.

What to do

• Download Fashion-MNIST (train/test sets)
• Downsample images to 8×8
• Apply PCA (k=16, 32, 64)
• Create T2 dataset (T-shirt vs Pullover)
• Create T4 dataset (T-shirt, Pullover, Coat, Shirt)
• Create T10 reduced dataset
• Save outputs as .npz in data/processed/
• Verify class balance and file contents

Where to work: data/preprocessing/