Total
33
Done
0
Progress
0%
Left
33
Est. time
~46 h
Level 1 — Geometric foundations
foundations 0/8 ▾Vectors: definition and operations
Addition, subtraction, scaling. Geometric meaning in ℝ² and ℝ³.
3Blue1Brown Ep.1 · Strang Ch.1
Dot product and norms
Dot product, vector length, angle between vectors, projection.
3Blue1Brown Ep.9 · Strang Ch.1
Matrices: definition and types
Square, identity, triangular and symmetric matrices. The transpose.
Strang Ch.1
Matrix multiplication
Why it is not commutative. Geometric intuition as composition of transformations.
3Blue1Brown Ep.3–4 · Strang Ch.2
Systems of linear equations
Matrix form Ax=b. Existence and uniqueness of solutions.
Strang Ch.1–2
Gaussian elimination
Row echelon form, row operations, elementary matrices.
Strang Ch.2
Matrix inverse
When it exists, how to compute it, and how it relates to Gauss-Jordan.
3Blue1Brown Ep.7 · Strang Ch.2
Basic implementation in NumPy
np.array, np.dot, np.linalg.inv, np.linalg.solve. Solve a real system.
NumPy docs — linalg module
Level 2 — Theoretical core
theory 0/9 ▾Vector spaces
Axiomatic definition. Subspaces, span, closure under addition and scaling.
Axler Ch.1–2 · Strang Ch.3
Linear independence
When a set of vectors is independent. Reducing a set to a basis.
Strang Ch.3 · Axler Ch.2
Basis and dimension
Basis of a space, dimension, coordinates in a basis.
Axler Ch.2 · Strang Ch.3
The four fundamental subspaces
Column space, null space, row space, left null space. Rank-nullity theorem.
Strang Ch.3 — one of the most important chapters
Linear transformations
Definition, kernel and image, matrix representation, change of basis.
Axler Ch.3 · 3Blue1Brown Ep.3
Determinants
Definition, properties, geometric meaning (volume), cofactors.
Strang Ch.5 · 3Blue1Brown Ep.6
Eigenvalues and eigenvectors
Characteristic equation and polynomial, multiplicity.
3Blue1Brown Ep.14 · Strang Ch.6
Diagonalization
When a matrix is diagonalizable. Change of basis to an eigenbasis.
Strang Ch.6 · Axler Ch.8
Symmetric matrices and the spectral theorem
Real eigenvalues, orthogonal eigenvectors, spectral decomposition.
Strang Ch.6 · Axler Ch.7
Level 3 — Decompositions and numerical linear algebra
numerical 0/8 ▾Orthogonality and projections
Orthogonal vectors, orthogonal complement, projection onto a subspace.
Strang Ch.4
Gram-Schmidt process
Building an orthonormal basis. QR factorization.
Strang Ch.4 · Axler Ch.6
Least squares
Overdetermined systems, the pseudoinverse, linear regression as algebra.
Strang Ch.4 — a direct link to regression
LU decomposition
Matrix factorization, solving systems efficiently, pivoting.
Strang Ch.2 · scipy.linalg.lu
SVD — Singular value decomposition
Σ, U, Vᵀ. Geometric meaning. The most powerful theorem in applied linear algebra.
Strang Ch.7 · "Learning from Data" Ch.1
PCA from SVD
Implement PCA by hand with SVD, without sklearn. Explained variance, dimensionality reduction.
numpy.linalg.svd — implement from scratch
Positive definite matrices
Definition, Sylvester's criterion, link to convexity and optimization.
Strang Ch.6 · "Learning from Data"
Matrix norms and conditioning
Frobenius norm, 2-norm, condition number, numerical stability.
Trefethen & Bau — Numerical Linear Algebra Ch.2
Level 4 — Applications by domain
applications 0/8 ▾Markowitz portfolio
Covariance matrix of returns, quadratic optimization, efficient frontier.
Fabozzi — Quantitative Equity Portfolio Management
Factor models (Fama-French)
Multiple regression as a linear system. Decomposing returns by factor.
Strang "Learning from Data" + the original papers
Risk component analysis
Eigenvalues of the correlation matrix to detect systemic risk.
Quantitative risk management
Backpropagation as matrix algebra
Gradients as Jacobians, the chain rule in high dimensions, forward/backward pass.
Strang "Learning from Data" Ch.VII · Stanford cs231n
Attention in transformers
Q, K, V as matrices. Attention score = softmax(QKᵀ/√d)·V. Why it works.
Attention Is All You Need (Vaswani et al.) + 3Blue1Brown LLM series
PageRank and graph algebra
A graph as an adjacency matrix, the dominant eigenvector, the power method.
Strang Ch.10 · Wikipedia PageRank
Lattice cryptography
A lattice as a discrete subgroup of ℝⁿ. SVP and CVP as hard problems. The basis of Kyber/NTRU.
Peikert — A Decade of Lattice Cryptography (2016, free)
LLL and basis reduction
The Lenstra-Lenstra-Lovász algorithm for finding "nearly orthogonal" bases of lattices.
Hoffstein et al. — An Introduction to Mathematical Cryptography
// bibliography
The books, courses and practice sites behind this roadmap.
// books
Introduction to Linear Algebra ↗
The main text. Most topics point to one of its chapters.
Linear Algebra Done Right ↗
Proof-based and operator-first. The theory after Strang.
Linear Algebra and Learning from Data ↗
SVD, PCA, optimization and deep learning as linear algebra.
Numerical Linear Algebra
Conditioning, stability and the algorithms behind np.linalg.
An Introduction to Mathematical Cryptography ↗
Lattices, LLL and the math behind post-quantum crypto.
// courses
MIT OCW 18.06SC — Linear Algebra ↗
Lectures, problem sets and exams. The backbone of levels 1–3.
MIT OCW 18.065 — Matrix Methods in Data Analysis and ML ↗
The course version of Learning from Data.
Essence of Linear Algebra ↗
Short videos for the geometric intuition. Watch before reading.
Computational Linear Algebra ↗
Code-first: NumPy, SVD and numerical stability in notebooks.
// practice
MIT 18.06 — problem sets & exams ↗
Weekly problem sets and past exams, with solutions.
Problems in Mathematics — Linear Algebra ↗
Hundreds of solved problems, tagged by topic.
numpy-100 ↗
100 NumPy exercises, from basic to hard.