Linear algebra · 04 · Many unknowns, one act · 9 min
Solving systems & why it scales
Two equations, two unknowns — algebra handles it. Two hundred thousand unknowns? That's Tuesday for a weather model. Linear algebra's final gift is making “solve them all at once” a single, scalable operation.
Build the intuition
A system is one matrix equation
The system x + 2y = 5, 3x − y = 1 packs into Ax = b: matrix of coefficients, vector of unknowns, vector of results. Solving means asking: which input vector does action A send to b? One equation, any number of unknowns inside.
The inverse: the undo button
If action A can be undone (determinant ≠ 0), the undo is itself a matrix, A⁻¹, and the solution is simply x = A⁻¹b: run the machine backwards. When det = 0 the machine flattened space — multiple inputs share outputs, and no clean undo exists. Geometry explains solvability.
Elimination: solving by tidying
In practice computers solve systems by Gaussian elimination: subtract multiples of rows to sweep variables out, column by column, until the answers fall out the bottom. It's the balance method from algebra, industrialized — and it's running inside every engineering tool you've heard of.
See it move
Solvability is geometry: verbs that keep space 2D can be undone; “Squash flat” has det → small — push it to zero and inputs become unrecoverable.
A worked example
Two foods, two targets
Rice has 4g protein and 45g carbs per serving; beans 9g and 27g. You want exactly 30g protein, 153g carbs.
System:
Eliminate: multiply the first by 3 (12r + 27b = 90), subtract from the second: 33r = 63… r ≈ 1.9, then b ≈ 2.5.
About 1.9 servings of rice, 2.5 of beans. Diet apps solve exactly such systems — with hundreds of foods and nutrients.
Out in the world
GPS is a solved system
Each satellite signal yields one equation relating your unknown position and clock error. Four satellites, four unknowns (x, y, z, t) — your phone solves the system in milliseconds, continuously, while you walk.
Common confusion, cleared
“More equations always pin the answer down better.”
Equations can be redundant (same info twice) or contradictory. What matters is independent information — which is exactly what determinants and rank measure.
“Computers solve systems by computing A⁻¹.”
Inverses are for understanding; elimination is for computing — it's faster and numerically safer. The concept and the algorithm divide the labor.
Check yourself
PracticeQuick check
A transformation has determinant 0. Solving Ax = b will…
The dot product of (1, 2) and (4, −2) is…
Recap
- Any linear system is one equation: Ax = b.
- Invertible action (det ≠ 0) → unique solution: run A backwards.
- Elimination scales the balance method to thousands of unknowns.