Chapter 1 — Key Takeaways

The big ideas

  • Linear algebra is the study of linear transformations of space. Determinants, eigenvalues, and the rest of the machinery exist to help us understand those transformations — they are tools, not the point. This is the book's first recurring theme, and everything in the next 39 chapters hangs off it.
  • A matrix is a function that transforms space. This is the single idea to carry forward. A $2\times 2$ matrix rotates, scales, shears, projects, or some combination — and its columns are the landing spots of the basis vectors (where east and north go). The grid of numbers is just how we write the transformation down. (Theme: geometry and algebra are two views of one object.)
  • Linearity = superposition. A transformation is linear when it preserves vector addition, $T(\mathbf{u}+\mathbf{v}) = T(\mathbf{u})+T(\mathbf{v})$, and scalar multiplication, $T(c\mathbf{v}) = c\,T(\mathbf{v})$. Combined: $T(c\mathbf{u}+d\mathbf{v}) = c\,T(\mathbf{u})+d\,T(\mathbf{v})$ — the whole is the sum of the scaled parts. A consequence: a linear map must fix the origin, $T(\mathbf{0})=\mathbf{0}$.
  • Knowing the map on the basis vectors determines it everywhere. Because of superposition, two columns of numbers pin down a transformation of all of infinite space. This is why the visualizer only has to track one unit square.
  • The determinant measures area scaling. Determinant 6 means areas grow sixfold; 1 means areas are preserved (rotations and shears); 0 means the transformation crushes a dimension and cannot be undone (it is singular); a negative value means orientation flips.
  • Linear algebra is everywhere, not just in physics. The same vectors-and-matrices appear in machine learning (layers = matrix multiplications), computer graphics (frames = matrix multiplications), quantum mechanics (states = vectors, gates = matrices), data science (datasets = matrices), signals (sounds/images = vectors), and economics (input–output models = linear systems). (Theme: linear algebra is the most applied branch of pure mathematics — learn it once, use it everywhere.)

Skills you gained

  • Stating the two rules of linearity and using superposition to compute a transformation's effect from its action on a few vectors.
  • Classifying an everyday process as linear or nonlinear with the two-question test (does adding inputs add outputs? does doubling the input double the output?).
  • Reading a $2\times 2$ matrix as a geometric motion by inspecting where its columns send the basis vectors.
  • Reading a transformation figure produced by visualize_2d and connecting the determinant in the title to the area change you see.
  • Recognizing the vectors, the matrix, and the transformation hiding inside a described real system.
  • Setting up the from-scratch toolkit/ package and running the canonical visualizer.

Terms to know

linear algebra · vector (an arrow / a list of numbers) · matrix (a function that transforms space; columns = images of basis vectors) · linear transformation · linearity · superposition · vector addition and scalar multiplication (the two preserved operations) · standard basis vectors $\mathbf{e}_1, \mathbf{e}_2$ · unit square · identity $I$ · scaling · rotation · shear · (orthogonal) projection · determinant (area-scaling factor) · singular (non-invertible, determinant 0) · affine (linear-plus-a-shift; not linear).

Notation introduced (locked for the whole book)

  • Vectors: bold lowercase, $\mathbf{v}, \mathbf{u}, \mathbf{x}$; columns by default.
  • Matrices: italic capital, $A, B, R, S, P$.
  • Components are 1-indexed in math ($v_1, a_{11}$) but 0-indexed in numpy (v[0], A[0,0]) — watch the shift whenever code appears.
  • Transpose is written $A^{\mathsf{T}}$ (introduced later); the matrix-multiply operator in numpy is @.

How this connects forward

  • Chapter 2 builds the vector rigorously (arrow ↔ list of numbers) and adds vectors.py to your toolkit.
  • Chapter 7 delivers the full payoff of "columns are images of basis vectors" — a matrix is a linear map.
  • Chapter 8 reveals matrix multiplication as composition of transformations and explains why order matters (non-commutativity) — the rule you should not memorize blindly.
  • Chapter 9 asks how to undo a transformation (the inverse), and why singular matrices (determinant 0, the projection case) can't be undone.
  • Chapter 11 makes the determinant-as-area-scaling idea precise.
  • Chapter 12 handles translations via homogeneous coordinates and builds the graphics pipeline (Case Study 2).
  • Chapter 21 studies rotations as orthogonal/unitary matrices — the gateway to quantum gates.
  • Chapters 29–33 cash in the seeded anchors: PageRank as a dominant eigenvector (Ch. 29), SVD image compression (Ch. 31), the recommender of Case Study 1 (Chs. 30, 33), and PCA (Ch. 32).
  • The qubit, seeded here, returns in Chapters 21, 27, and 34.

The recurring anchor

You built visualize_2d, the 2D transformation visualizer. It returns in nearly every geometric chapter — scaling and shear in Chapters 7–8, area = determinant in Chapter 11, change of basis in Chapter 16, rotations in Chapter 21, invariant directions = eigenvectors in Chapter 23, and rotate–stretch–rotate = SVD in Chapter 30. The moving unit square is the thread that ties Chapter 1 to Chapter 40. Whenever you meet a new matrix, ask the question that organizes this whole subject: what does it DO to space?