STUDY GUIDE · 2 INTERACTIVE LABS · ABOUT 12 MIN

A matrix isn't a table of numbers. It's a way to move space.

Every 3D game, search engine, recommendation system and neural network runs on linear algebra. At its heart is one idea: a matrix transforms vectors, stretching, rotating or squashing the space they live in. Drag one and see.

3D graphicsEvery frame rotates and projects millions of points with matrices.
Search rankingGoogle's original PageRank was an eigenvector of the web's link matrix.
Neural networksEach layer is a matrix multiplication plus a twist.
Image compressionKeep the biggest singular values, drop the rest.

What is linear algebra?

Linear algebra is the mathematics of vectors, matrices and the linear transformations between them. It studies systems of linear equations, how transformations stretch and rotate space, and the special directions (eigenvectors) that a transformation leaves pointing the same way.

A transformation is "linear" if it keeps grid lines straight, parallel and evenly spaced, and leaves the origin in place. That's why a 2×2 matrix is fully described by where it sends just two arrows: î and ĵ.

VectorA quantity with size and direction: an arrow from the origin.
MatrixA grid of numbers that describes a linear transformation.
DeterminantHow much the transformation scales area (or volume).
EigenvectorA direction the transformation only stretches, never turns.
LAB 1 · MATRICES AS TRANSFORMATIONS

Watch a matrix reshape the grid

The first column of the matrix is where î (red) lands; the second is where ĵ (green) lands. Everything else follows. The shaded square shows how areas scale, and the dashed lines are eigenvector directions, which are only stretched, never turned.

[
1.01.00.01.0
]
1.00determinant (area scale)
1.00eigenvalues
Determinant 1: shapes change but areas stay exactly the same, as in a shear or rotation.
LAB 2 · SYSTEMS OF EQUATIONS

Two equations, two lines, one answer, or none

Each linear equation in x and y is a line. Solving both at once means finding where they cross. If the coefficient matrix has determinant zero, the lines are parallel: no solution, or infinitely many if they coincide.

x + y = 3Line 1 is fixed.
x − y = 1
-2determinant
x = 2, y = 1unique solution
det = a₁b₂ − a₂b₁ = -2 ≠ 0, so the lines cross exactly once. Cramer's rule gives the solution directly from determinants.

Core ideas at a glance

Matrix × vector[a b; c d]·(x, y) = (ax + by, cx + dy)Each output is a weighted sum.
Determinant (2×2)det = ad − bc0 means the matrix squashes space and has no inverse.
Inverse (2×2)A⁻¹ = (1/det)·[d −b; −c a]Undoes the transformation.
Eigen equationA·v = λ·vv keeps its direction; λ is the stretch factor.
Characteristic equationλ² − (a + d)λ + det = 0Solve it to find a 2×2 matrix's eigenvalues.
Matrix multiplication(AB)v = A(Bv)Apply B first, then A. Order matters: AB ≠ BA in general.
Dot productu·v = |u||v| cos θZero when vectors are perpendicular.
RankNumber of independent columnsThe dimension of the space the matrix can reach.

From Chinese counting rods to Google

c. 200 BCThe Nine Chapters.A Chinese text solved systems of equations with counting rods, using what we now call Gaussian elimination.
1693Determinants.Leibniz used them to solve equations, as did Seki Takakazu in Japan around the same time.
1810Gaussian elimination.Gauss used systematic elimination to calculate the orbit of the asteroid Pallas.
1858Matrix algebra.Arthur Cayley treated matrices as objects you can add, multiply and invert.
1888Vector spaces.Giuseppe Peano gave the modern abstract definition.
1998PageRank.Brin and Page ranked the web by computing the dominant eigenvector of a giant link matrix.

Further reading

Introduction to Linear AlgebraGilbert Strang · with free MIT lectures
Linear Algebra Done RightSheldon Axler · the abstract, proof-based route
Essence of Linear Algebra3Blue1Brown · free animated video series
Test yourself: 20 linear algebra questionsMachine learning guide

Picturing transformations is spatial reasoning. Test yours.

Our full IQ test covers spatial, logical, numerical and verbal reasoning: 60 questions, about 30 minutes.

Take the full IQ test