BS Math and BSCS students can use this guide to study Interactive Linear Algebra by Dan Margalit and Joseph Rabinoff of the Georgia Institute of Technology. It is free under the GNU Free Documentation License, Version 1.2 or any later version. The PDF listed here is dated June 3, 2019, and it covers a complete first course in linear algebra: systems of equations, vectors and subspaces, linear transformations, determinants, eigenvalues and eigenvectors, and orthogonality with least squares.
What makes this book different is how it is organised. Every section opens with a list of objectives, and each idea is labelled as a recipe, a vocabulary word, a theorem or a picture, so you always know whether to practise it, memorise it or draw it. The online edition adds over 140 interactive demos, and the PDF marks each one with a link. The text is about half computational and half conceptual, which suits the way BS Math and BSCS linear algebra exams mix row reduction with questions about span, rank, eigenvalues and projections.
Book Overview
| Course | Linear Algebra |
| Degree Programs | BS Math, BSCS, Engineering |
| Level | University — first course in linear algebra; it starts from solving systems of equations and introduces R^n from scratch, with complex numbers reviewed in Appendix A |
| Edition | Version dated June 3, 2019 (copyright line 2017); the official PDF and online edition carry the same date |
| Authors | Dan Margalit and Joseph Rabinoff, Georgia Institute of Technology (Larry Rolen wrote many of the exercises) |
| Publisher | Georgia Institute of Technology, School of Mathematics |
| Structure | 6 chapters and 32 sections, plus 4 appendices, in a 455-page PDF |
| Exercises | None printed in this PDF; Appendix C (Hints and Solutions to Selected Exercises) is a heading page only |
| Language | English |
| License | GNU Free Documentation License (GFDL), Version 1.2 or any later version, with no Invariant Sections, no Front-Cover Texts and no Back-Cover Texts |
| Format | Free PDF, plus a free online edition with over 140 interactive demos |
Chapter List
Chapter 1: Systems of Linear Equations: Algebra
Difficulty: Easy · Key topics: linear equations, solution sets, R^n, augmented matrix, row operations, echelon form, pivots, free variables, parametric form
The book opens with traffic flow, chemical reactions, rabbit populations and asteroid orbits to show why linear systems matter. Section 1.1 defines linear equations and solution sets, uses R^n to label points on a line, plane or higher-dimensional space, and draws what solution sets look like. Section 1.2 turns the elimination method into row reduction of an augmented matrix, defines row echelon and reduced row echelon form, and proves that every matrix is row equivalent to exactly one reduced form. Section 1.3 uses free variables to write every solution in parametric form and sorts systems into no solution, one solution or infinitely many.
Key Points:
- A system is consistent if it has at least one solution and inconsistent otherwise; the solution set is the collection of all solutions.
- The three row operations (replace a row, scale a row, swap two rows) never change the solution set.
- Every matrix is row equivalent to one and only one matrix in reduced row echelon form.
- A pivot is the first nonzero entry of a row in echelon form; a free variable is a variable whose column has no pivot.
- Parametric form writes each pivot variable in terms of the free variables, so one equation in three unknowns describes a plane in R^3.
- If the last column is a pivot column the system has no solution; if every other column has a pivot there is exactly one solution; otherwise there are infinitely many.
Memory Tip: Read the pivots, not the numbers. A pivot in the augmented column means no solution, a pivot in every variable column means one solution, and any variable column without a pivot means infinitely many.
Common Mistake: Treating a free variable as something left unsolved. It is a parameter you may set to any number, and every other variable is then determined by it.
Important Questions:
- What is a free variable? In a consistent system, a variable whose column has no pivot in the row echelon form. It can take any value, and the pivot variables are expressed through it.
- How many solutions can a linear system have? Zero, exactly one, or infinitely many. Which case occurs is read from the pivot positions of the reduced augmented matrix.
Chapter 2: Systems of Linear Equations: Geometry
Difficulty: Medium · Key topics: vectors, linear combinations, span, Ax = b, homogeneous systems, parametric vector form, linear independence, subspaces, basis, dimension, coordinates, rank theorem
This chapter asks two questions about Ax = b: which x solve it, and which b make it solvable at all. Sections 2.1 to 2.3 build the language, namely vectors and linear combinations, spans, and the matrix-vector product, so a system can be written four equivalent ways. Section 2.4 shows that a homogeneous solution set is a span and that any consistent solution set is a translate of one. Sections 2.5 to 2.8 add linear independence, subspaces, bases, dimension and coordinate systems. Section 2.9 closes the chapter with the rank theorem, which links the two questions.
Key Points:
- Ax is the linear combination of the columns of A with the entries of x as weights, so Ax = b is consistent exactly when b lies in the span of the columns.
- The solution set of Ax = 0 is a span; the solution set of a consistent Ax = b is one particular solution plus that span.
- Vectors are linearly independent when the only dependence relation is the trivial one, which happens exactly when the matrix of those columns has a pivot in every column.
- A subspace contains 0 and is closed under addition and scalar multiplication; the column space Col(A) and null space Nul(A) are the standard examples.
- A basis is a linearly independent set that spans a subspace; the dimension is the number of vectors in any basis.
- The pivot columns of A, taken from the original matrix, form a basis of Col(A), and the free variables give a basis of Nul(A).
- Rank theorem: rank(A) + nullity(A) = the number of columns of A.
Memory Tip: Keep two pictures apart. The solution set lives in the input space (which x work for one b); the column space lives in the output space (which b work at all). The rank theorem is the trade-off between them.
Common Mistake: Row reducing a matrix and then reading a basis of its column space off the reduced matrix. Use the reduced form only to find which columns are pivot columns, then take those columns of the original matrix.
Important Questions:
- When does Ax = b have a solution for every b in R^m? Exactly when A has a pivot in every row, which is the same as saying the columns of A span R^m.
- What does the rank theorem say? For a matrix with n columns, dim Col(A) + dim Nul(A) = n; that is, the number of pivot columns plus the number of free variables equals the number of columns.
Chapter 3: Linear Transformations and Matrix Algebra
Difficulty: Medium · Key topics: matrix transformations, domain, codomain, range, one-to-one, onto, linear transformations, standard matrix, matrix multiplication, inverses, invertible matrix theorem
Chapter 3 reads b = Ax as a function. Section 3.1 introduces domain, codomain and range and pictures familiar 2 × 2 examples: projection, reflection, dilation, rotation and shear. Section 3.2 turns one-to-one and onto into pivot conditions. Section 3.3 defines linear transformations and shows each one is a matrix transformation whose columns are the images of the standard coordinate vectors. Matrix multiplication then appears as composition of transformations (3.4), inverses are computed by row reduction (3.5), and Section 3.6 gathers many equivalent conditions into the invertible matrix theorem.
Key Points:
- A matrix transformation T(x) = Ax has domain R^n, codomain R^m and range Col(A); the range need not be the whole codomain.
- T is one-to-one exactly when the columns of A are linearly independent (a pivot in every column) and onto exactly when they span R^m (a pivot in every row).
- A transformation is linear if T(u + v) = T(u) + T(v) and T(cu) = cT(u); its matrix has columns T(e1), …, T(en).
- The product AB is the matrix of the composition that applies B first and then A; in general AB is not equal to BA.
- An inverse is found by row reducing (A | I) to (I | A^(-1)), and (AB)^(-1) = B^(-1) A^(-1).
- Invertible matrix theorem: A is invertible exactly when it has n pivots, Nul(A) = {0}, its columns are independent, its columns span R^n, and Ax = b has a unique solution for every b.
Memory Tip: To build the matrix of any linear transformation, feed it the standard basis vectors and use the outputs as columns. For a composition, read right to left: AB means do B first.
Common Mistake: Writing (AB)^(-1) = A^(-1) B^(-1). The order flips, because undoing a composition means undoing the last step first.
Important Questions:
- What is the difference between one-to-one and onto for a matrix transformation? One-to-one means different inputs give different outputs (independent columns). Onto means every vector of the codomain is an output (columns span R^m).
- How do you find the inverse of a square matrix? Row reduce the augmented matrix (A | I). If the left half becomes I, the right half is A^(-1); if not, A is not invertible.
Chapter 4: Determinants
Difficulty: Medium · Key topics: determinant, row operations, triangular matrices, transpose, cofactor expansion, adjugate, Cramer’s rule, parallelepiped, volume
The chapter opens by placing determinants in the second half of the book’s plan, the eigenvalue problem. Section 4.1 defines the determinant by how it reacts to row operations (a replacement changes nothing, scaling a row by c scales it by c, a swap flips the sign, and the identity has determinant 1), then derives the invertibility test, multiplicativity and det(A^T) = det(A). Section 4.2 gives the recursive cofactor expansion along any row or column, with Cramer’s rule and an adjugate formula for the inverse. Section 4.3 explains the geometry: |det A| is the volume of a parallelepiped and the factor by which A scales volume.
Key Points:
- Row replacement leaves det unchanged, scaling a row by c multiplies it by c, swapping two rows multiplies it by −1, and det(I) = 1.
- A square matrix is invertible exactly when det(A) is nonzero, and then det(A^(-1)) = 1/det(A).
- det(AB) = det(A) det(B) and det(A^T) = det(A), so column operations work just like row operations.
- The determinant of a triangular matrix is the product of its diagonal entries.
- Cofactor expansion works along any row or column with a checkerboard sign pattern; pick the one with the most zeros.
- |det(A)| is the volume of the parallelepiped spanned by the columns, and vol(T(S)) = |det(A)| vol(S) for any region S.
Memory Tip: Think of det as a signed volume-scaling factor. Zero means the transformation flattens space and cannot be undone; a negative sign means orientation is flipped.
Common Mistake: Assuming det(A + B) = det(A) + det(B). The determinant respects products, not sums.
Important Questions:
- How does the determinant tell you whether a matrix is invertible? A square matrix is invertible exactly when det(A) is nonzero; a zero determinant means the rows (and columns) are linearly dependent.
- What does |det(A)| measure geometrically? The volume of the parallelepiped spanned by the columns of A, which is also the factor by which the transformation x to Ax scales volumes.
Chapter 5: Eigenvalues and Eigenvectors
Difficulty: Hard · Key topics: eigenvectors, eigenvalues, eigenspaces, characteristic polynomial, trace, similarity, diagonalization, multiplicity, complex eigenvalues, rotation-scaling, stochastic matrices, PageRank
Starting from the rabbit-population example, the chapter treats Ax = λx. Section 5.1 defines eigenvectors, eigenvalues and eigenspaces and extends the invertible matrix theorem. Section 5.2 finds eigenvalues as roots of the characteristic polynomial det(A − λI). Section 5.3 introduces similar matrices, A = C B C^(-1), as the same transformation seen in a different coordinate system. Section 5.4 asks when a matrix is similar to a diagonal one and covers multiplicity and fast powers. Section 5.5 handles complex eigenvalues through rotation-scaling matrices, and Section 5.6 applies the theory to stochastic matrices, steady states and Google’s PageRank.
Key Points:
- An eigenvector is a nonzero v with Av = λv; λ is an eigenvalue when this has a nontrivial solution, and eigenvectors for distinct eigenvalues are linearly independent.
- Eigenvalues are the roots of det(A − λI); for a 2 × 2 matrix this is λ^2 − Tr(A)λ + det(A).
- Similar matrices (A = C B C^(-1)) share the characteristic polynomial (and so the eigenvalues, trace and determinant), but sharing it does not make two matrices similar.
- A is diagonalizable exactly when it has n linearly independent eigenvectors; then A = C D C^(-1) and A^k = C D^k C^(-1).
- 1 ≤ geometric multiplicity ≤ algebraic multiplicity, and a strict inequality (as for a shear) blocks diagonalization.
- A real 2 × 2 matrix with a complex eigenvalue is similar to a rotation-scaling matrix that scales by |λ| and rotates by the angle of λ.
- A positive stochastic matrix has a unique steady state that its powers converge to (Perron-Frobenius), which is the idea behind PageRank.
Memory Tip: An eigenvector is a direction the matrix does not turn, and λ says how much it stretches along that direction. Diagonalizing means switching to a coordinate system built entirely from such directions.
Common Mistake: Assuming a repeated eigenvalue is harmless. Count the independent eigenvectors: a shear has a repeated eigenvalue but only a line of eigenvectors, so it is not diagonalizable.
Important Questions:
- How do you find the eigenvalues and eigenvectors of a square matrix? Solve det(A − λI) = 0 for λ, then for each root find a basis of the null space of A − λI, which is the λ-eigenspace.
- When is a matrix diagonalizable? Exactly when it has n linearly independent eigenvectors, that is, when the eigenspace dimensions add up to n. Having n distinct eigenvalues is enough.
Chapter 6: Orthogonality
Difficulty: Medium · Key topics: dot product, length, orthogonal complement, row space, orthogonal projection, orthogonal sets, Gram-Schmidt, least squares, normal equations, best-fit problems
The final chapter handles the third goal: when Ax = b has no solution, find the best approximate one. Section 6.1 defines the dot product, length, distance and orthogonality. Section 6.2 builds orthogonal complements and proves that row rank equals column rank. Section 6.3, the core of the chapter, shows that every vector splits uniquely into a part in a subspace W and a part in W⊥, and that the first part is the closest vector in W. Section 6.4 uses orthogonal sets and the Gram-Schmidt process to simplify projections. Section 6.5 derives least squares and applies it to best-fit lines, parabolas, planes, trigonometric functions and ellipses.
Key Points:
- The dot product gives length, ||x|| = sqrt(x · x), and distance; x and y are orthogonal exactly when x · y = 0.
- The orthogonal complement W⊥ of a subspace W satisfies dim W + dim W⊥ = n, and Col(A)⊥ = Nul(A^T).
- Row rank equals column rank, so A and A^T always have the same number of pivots.
- Every x splits uniquely as x_W + x_W⊥, and x_W, the orthogonal projection, is the closest vector of W to x.
- A set of nonzero orthogonal vectors is linearly independent, and Gram-Schmidt converts any basis into an orthogonal one.
- Least-squares solutions of Ax = b are exactly the solutions of A^T A x = A^T b.
Memory Tip: Closest point means perpendicular drop. The error vector b − A(x-hat) must be orthogonal to the column space, and that one sentence produces the normal equations.
Common Mistake: Projecting onto the span of several vectors by adding up one-line projections when the vectors are not orthogonal. That shortcut only works for an orthogonal basis; otherwise run Gram-Schmidt first or use A^T A.
Important Questions:
- What is a least-squares solution of Ax = b? A vector x-hat for which Ax-hat is as close to b as possible, meaning Ax-hat equals the orthogonal projection of b onto Col(A). It is found by solving A^T A x = A^T b.
- What is the orthogonal projection of x onto a subspace W? The unique vector x_W in W for which x − x_W lies in W⊥; it is the closest vector in W to x.
Appendices A-D: Complex Numbers, Notation, Hints, GNU Free Documentation License
Appendix · Difficulty: Easy · Key topics: complex arithmetic, conjugates, absolute value, fundamental theorem of algebra, notation table, licence text
Appendix A reviews complex numbers, motivated by a rotation matrix that has no real eigenvalues. It covers addition, multiplication using i^2 = −1, the conjugate, absolute value and division, then states the fundamental theorem of algebra and shows that complex roots of real polynomials come in conjugate pairs, with quadratic and cubic examples. Appendix B is a symbol table listing each notation with the page where it first appears. Appendix C is titled Hints and Solutions to Selected Exercises, but in this PDF it is a heading page with nothing printed under it. Appendix D reproduces the licence text.
Key Points:
- The conjugate of a + bi is a − bi, and z times its conjugate equals |z|^2, a nonnegative real number.
- Fundamental theorem of algebra: a polynomial of degree n has exactly n complex roots, counted with multiplicity.
- Complex roots of a polynomial with real coefficients come in conjugate pairs, so a real cubic has three real roots or one real root and a conjugate pair.
- The notation table (Col, Nul, rank, nullity, det, Tr, adj, W⊥ and others) is a quick way to look up a symbol you forgot.
- Appendix C holds no hints in this PDF, and Appendix D prints the GNU Free Documentation License, Version 1.3.
Memory Tip: A real matrix with a complex eigenvalue always brings its conjugate along, so for 2 × 2 matrices you only ever need to analyse one of the pair.
Common Mistake: Dividing by a complex number by cancelling i. Multiply the top and bottom by the conjugate of the denominator instead, so the denominator becomes a real number.
Important Questions:
- How do you divide one complex number by another? Multiply numerator and denominator by the conjugate of the denominator; the denominator then becomes the real number |w|^2.
- What does the fundamental theorem of algebra say? Every polynomial of degree n factors into n linear factors over the complex numbers, so it has exactly n complex roots counted with multiplicity.
How to Study This Book
The book frames the whole course as three goals: solve Ax = b, solve Ax = λx, and approximately solve Ax = b. In this PDF, Chapters 1 to 3 handle the first goal, Chapters 4 and 5 the second (determinants exist mainly to find eigenvalues), and Chapter 6 the third. The authors say the text contains exactly what Georgia Tech’s Math 1553 covers, so there is no optional filler. A full BS Math or BSCS linear algebra course can read it cover to cover in order.
Chapters 1 and 2 are the foundation, and Chapter 2 is where most students slow down. Spend extra time on Section 2.4 (the shape of a solution set), then take Sections 2.5 to 2.7 (independence, subspaces, basis and dimension) one at a time. Every later chapter keeps asking whether vectors are independent, whether they span, and how many pivots a matrix has.
Use the Objectives box at the top of each section as a checklist. The book sorts its ideas into recipes (procedures to practise by hand until they are automatic), vocabulary, essential vocabulary (definitions you must be able to state exactly, such as span, linearly independent, basis, determinant and eigenvector), theorems and pictures. Knowing which kind of idea you are looking at tells you whether to practise it, memorise it or draw it.
Keep a one-page list of the invertible matrix theorem from Section 3.6. The book adds two more equivalent conditions in Section 5.1 (the determinant is nonzero, and 0 is not an eigenvalue), and this theorem is the quickest way to move between pivots, independence, spanning and invertibility in exam questions. Section 3.4 deserves care too, because matrix multiplication is introduced as composition of transformations.
In Chapter 4, Sections 4.1 and 4.2 are the computational core: row-reduction and cofactor methods. Section 4.3 explains why determinants behave as they do and is worth reading before multivariable calculus, where the same factor appears in change of variables. In Chapter 5, treat Sections 5.1 to 5.4 as essential. Read Appendix A before Section 5.5, because complex eigenvalues assume you are comfortable with conjugates and absolute values. Section 5.6 (stochastic matrices and PageRank) is an application, but a very good one for BSCS students.
In Chapter 6, read Sections 6.1 to 6.3 in order, then jump to Section 6.5 for least squares. Section 6.4 (orthogonal sets and Gram-Schmidt) is what makes projection calculations manageable, so return to it when the arithmetic in 6.3 or 6.5 gets heavy.
Use the online interactive demos. They are most helpful for spans (2.2), solution sets (2.4), matrix transformations (3.1), similarity and diagonalization (5.3 and 5.4) and complex eigenvalues (5.5), where seeing a point move makes the algebra clear. The PDF marks each demo with a link to the online version, and the book recommends a modern browser.
Know what the book leaves out. This PDF has no exercise sets, and its Appendix C hints page is empty, so pair it with problem sheets from your course or another text. It works only in R^n: there are no abstract vector spaces, inner product spaces, symmetric matrices, spectral theorem, singular value decomposition or Jordan form. If your course covers those, you will need a second book.
Used In These Programs
This book is used for the Linear Algebra course in: BS Mathematics · BS Computer Science, and in the linear algebra component of engineering degrees. Browse all Linear Algebra books or all Mathematics books.
Who Should Read This
This book suits a BS Math, BSCS or engineering student in a first university course in linear algebra, especially one that stresses computation and geometry over formal proof. It starts from systems of equations, builds R^n from scratch, and keeps returning to pictures, so it is a comfortable choice for a student who finds proof-first texts intimidating. It is also a good revision text before an exam, because every section lists what you must be able to do. It is less suited to students whose course covers abstract vector spaces, inner product spaces, the spectral theorem or the singular value decomposition, and it has no exercise sets in this PDF. For a proof-based treatment, pair it with Jim Hefferon’s Linear Algebra or Robert Beezer’s A First Course in Linear Algebra. Students heading toward data science or machine learning will find the eigenvalue and least-squares chapters directly useful.
Applicable Universities
This book is useful for students at Pakistani universities offering BS Math, BSCS and engineering programmes including Punjab University, Virtual University, COMSATS, FAST, UET, NUST, and other HEC-recognized institutions.
Official Copy of Interactive Linear Algebra
This book is free from its official source. Click below to open the official Interactive Linear Algebra page at Georgia Tech, where you can download the PDF version or read the full online edition with its interactive demos.
→ Visit Official PageFAQs
Is Interactive Linear Algebra free?
Yes. The book is published under the GNU Free Documentation License, Version 1.2 or any later version, with no Invariant Sections, no Front-Cover Texts and no Back-Cover Texts. You can download the PDF or read the online edition at no cost, and the licence also lets you copy, redistribute and modify the text under the same terms.
Which edition is this?
The PDF and the online edition are both dated June 3, 2019, and the copyright line reads 2017. The book has no numbered editions. The PDF currently offered on the official site matches it in both size and date, and the preface says a variant fine-tuned for Georgia Tech’s Math 1553 course also exists. This page describes the full (master) version.
Does it include solved exercises?
No. The PDF contains no exercise sets, and Appendix C, titled Hints and Solutions to Selected Exercises, is a heading page with nothing printed under it. Plan to pair the book with problem sheets from your course or with another text that has exercises.
Is it good for complete beginners?
Yes, with one caution. It starts from solving systems of equations and builds vectors, subspaces and matrices step by step, with many pictures. It is compact, though, and Chapter 2 introduces independence, subspaces and bases quickly, so a student new to this style of mathematics should work through each section’s examples rather than skim.
Does it cover eigenvalues, diagonalization and least squares?
Yes. Chapter 5 covers eigenvectors, the characteristic polynomial, similarity, diagonalization, complex eigenvalues and stochastic matrices, and Chapter 6 covers orthogonal projection, Gram-Schmidt and the method of least squares. It does not cover abstract vector spaces, symmetric matrices, the spectral theorem or the singular value decomposition, because it works only in R^n.
Do I need the online version to use the demos?
The demos run only in the online edition, in a modern browser such as Chrome, Firefox or Safari. The PDF is fully readable without them, and it marks each demo with a link to the online page. The book says it contains over 140 interactive demos, which are most useful for spans, solution sets, matrix transformations and eigenvectors.
Related Books
- Linear Algebra – Jim Hefferon (4th Edition)
- A First Course in Linear Algebra – Robert Beezer
- Mathematics for Machine Learning – Deisenroth, Faisal & Ong
- Calculus Volume 3 – Strang & Herman (OpenStax)
Interactive Linear Algebra is a compact, geometry-first course that moves from row reduction to eigenvalues and least squares in six chapters, with interactive demos online to make the pictures concrete. It works best as a main text for a computational first course or as revision material before an exam. Browse more Mathematics books for the rest of your semester.
Interactive Linear Algebra by Dan Margalit and Joseph Rabinoff, Georgia Institute of Technology (PDF dated June 3, 2019; copyright 2017 Georgia Institute of Technology). Free under the GNU Free Documentation License, Version 1.2 or any later version, with no Invariant Sections, no Front-Cover Texts and no Back-Cover Texts. Official source: https://textbooks.math.gatech.edu/ila/
© Copyright Policy
Freebooks.pk is an educational resource for students in Pakistan. Textbooks and other books belong to their respective boards, authors and publishers, and all rights remain with them. If you are the copyright holder of any material on this page and want it changed or removed, please contact us or file a request under our DMCA policy. We review every request and remove reported content promptly.
