What a Matrix Actually Is, and Why Programmers Should Care
A matrix is a spreadsheet with a rulebook for combining it with other spreadsheets — the intimidating part is the notation, not the idea.
Imagine an ordinary spreadsheet: rows and columns of numbers, nothing more. That's a matrix. Every intimidating thing said about matrices in a textbook is built on top of that one plain idea — a grid of numbers, arranged in rows and columns, that you can do specific, well-defined things to.
The size of a grid gets described as "rows by columns" — a matrix with 2 rows and 3 columns is called a 2×3 matrix. A single row or a single column is still a matrix, just a especially narrow one; that special case is usually called a vector, which is its own article on this site.
Programmers run into matrices constantly, often without the word "matrix" ever showing up in the code. A digital image is a grid of pixel brightness or color values — a matrix. A 3D game's engine tracks every object's position, rotation, and scale using matrices, because a matrix can represent "move, spin, and stretch this object" as a single combinable unit. A spreadsheet-style table in a database query result is, structurally, the same grid shape.
The one genuinely new idea a matrix introduces is that grids combine with their own specific rule, called matrix multiplication, and it isn't like multiplying two spreadsheet cells together one at a time. Combining two matrices means each value in the result is built from an entire row of one matrix and an entire column of the other, matched up position by position and added together. It looks unfamiliar the first time, but it's a fixed, mechanical recipe — not a judgment call — which is exactly why a computer can do millions of these combinations a second without getting confused.
Here's a 2×3 matrix, exactly as plain as it looks:
Once a matrix reads as "a grid, plus a specific rule for combining grids" instead of a wall of bracket notation, the rest of linear algebra gets a lot less intimidating — vectors, transformations, and eigenvectors, covered elsewhere in this category, are all built directly on this one idea.
[ 1 2 3 ]
[ 4 5 6 ]Two rows, three columns — that's the entire definition. The numbers themselves could be pixel values, coordinates, or anything else that fits a grid.