Matrix
A rectangular grid of numbers arranged in rows and columns.
- Represents data tables AND linear transformations.
- Transpose, inverse, rank, eigenvalues describe its action.
- GPUs accelerate ML because they accelerate matrix math.
When to use: Datasets, neural-network layers, attention weights - anywhere data has 2D structure.
Example: 100×5 matrix = 100 customers × 5 features. 768×768 = one self-attention projection in a small transformer.