1. Linear Systems of Equations
Definition 1.1 – Linear system
A list of unknowns and equations in which each unknown is only scaled by a constant and the scaled unknowns are only added — no exponents, no products of unknowns, no fancy functions.
Standard layout: unknowns on the left, lined up vertically (insert 0 coefficients where needed), constants on the right.
Example.
Geometric meaning
is a linear transformation, so solving
means: find the vector that lands on after the transformation.
2. Case : The Inverse
When space is not squished to lower dimension, exactly one vector lands on . Find it by playing the transformation in reverse and following back.
Definition 2.1 – Inverse matrix
The reverse-playback is itself a linear transformation, the inverse : the unique transformation such that doing then returns everything to where it started.
Solution
| Rotate 90° counter-clockwise | Rotate 90° clockwise |
| Shear pushing one unit right | Shear pushing one unit left |
Worked example (2×2).
Check: ✓, ✓.
For a randomly chosen square matrix, is by far the most likely case: equations in unknowns usually have a unique solution.
3. Case : No Inverse
If squishes space into a lower dimension, there is no inverse: you can’t “unsquish” a line back into a plane — that would require sending one input to a whole line of outputs, which a function can’t do.
A solution can still exist — but only if happens to lie on the squished output (e.g. on that line).
⚠️ Pitfall: means “no inverse”, not “no solution”. It means either no solution or infinitely many.
4. Rank and Column Space
Definition 4.1 – Rank
The rank of a transformation is the number of dimensions of its output.
| Matrix | Output | Rank |
|---|---|---|
| 2×2, | Whole plane | 2 |
| 2×2, squishes to a line | Line | 1 |
| 3×3, | All of 3D | 3 |
| 3×3, squishes to a plane | Plane | 2 |
| 3×3, squishes to a line | Line | 1 |
For a 3×3 matrix, rank 1 is a more severe collapse than rank 2 (solutions are much less likely to exist), even though both have .
Definition 4.2 – Column space
The set of all possible outputs . Since the columns are where the basis vectors land:
So more precisely: rank = dimension of the column space.
Definition 4.3 – Full rank
Rank equal to the number of columns (the maximum possible).
- is always in the column space (linear transformations fix the origin).
- For full-rank transformations, only lands on .
5. Null Space (Kernel)
Definition 5.1 – Null space
The set of all vectors that land on the zero vector:
Also called the kernel.
| Transformation squishes… | Null space is… |
|---|---|
| 2D → a line | A line (in a different direction) |
| 3D → a plane | A line |
| 3D → a line | A plane |
| Nothing (full rank) | Just |
Connection to systems
When , the null space is exactly the set of all solutions of .
Worked example. , .
- Column space: the line spanned by (rank 1).
- Null space: solve → the line spanned by .
- has infinitely many solutions (it’s on the column line); has none.
6. Summary
| Tool | What it tells you |
|---|---|
| Inverse | If it exists, the unique solution |
| Column space | Whether a solution exists at all (?) |
| Null space | What the full set of solutions looks like |
| Rank | How severely space collapses |
Exercises: For , find the rank, describe the column space and null space geometrically, and decide whether is solvable (lesson page).