← all subjects

Linear Algebra

Linear algebra is the study of linear maps and the spaces they act on. Vectors are the objects; a matrix is the map — and almost every question in the subject is really a question about what that map does to space.

Read a matrix as a verb, not a table. Its columns record exactly where the basis vectors land, and that alone fixes the fate of every other vector — the grid deforms with them.

The first hard question is when a system Ax = b has a solution at all. Gaussian elimination answers it mechanically: clear each column below its pivot and the matrix steps into row-echelon form, where the solutions can simply be read off.

FIG_002
211112408070
Gaussian elimination: entries below each pivot fall to zero, stepping the matrix to row-echelon form.

Beneath the mechanics sit the vector spaces that organize the answers. Independence, span, and basis measure how many directions a set of vectors really reaches, and dimension counts them.

FIG_003
uv
αu+βv\alpha u + \beta v
Span: every scalar combination of two independent vectors sweeps out the whole plane they generate.

Some directions survive a transformation untouched in orientation. An eigenvector is carried straight back onto its own line, merely stretched by an eigenvalue — the axes along which the map acts most simply.

FIG_004
λ1v1\lambda_1 v_1
λ2v2\lambda_2 v_2
x
Eigenvectors: a generic vector is rotated, but these special directions are only scaled by λ.
FIG_001
ij
A=[10.60.30.9]A=\begin{bmatrix}1 & 0.6\\[2pt]0.3 & 0.9\end{bmatrix}
A matrix as a transformation: the square grid shears into parallelograms, carrying the basis vectors to its columns.

Orthogonality turns geometry into computation. Projecting a vector onto a subspace drops a perpendicular to the closest point in it — the idea behind least squares, orthonormal bases, and the singular value decomposition.

FIG_005
b
pp
span{a}
Orthogonal projection: the foot p is the closest point on the line to b, met at a right angle.

Determinants measure how a map scales volume and whether it is invertible; an eigenbasis diagonalizes it; and the SVD factors any matrix into a rotation, a set of stretches, and another rotation.

The payoff is a single language for systems of equations, geometry, and data — and a set of factorizations that make each of them computable, even in floating point.

Contents.

·
Progress.░░░░░░░░░░░░░░░░░░
Articles done:0 / 49
Complete:0%
Notes written:0
Highlights:0