Solving Linear Systems

01Row Picture, Column Picture & Matrix Form 02Row Picture vs Column Picture in 3D 03Singular vs Non-Singular Matrices 04Two Ways to Compute Ax Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

01Gaussian Elimination & Back Substitution 02Elimination Matrices 03Permutation Matrices and Inverses Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

015 Views of Matrix Multiplication 02Matrix Inverses and Singularity 03Gauss-Jordan Elimination for Inverses Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01Product Inverse & Transpose Rules 02A = LU Factorization 03Gaussian Elimination: ⅓n³ Operation Count 04Permutation Matrices: P⁻¹ = Pᵀ Problem set0/10 Problem set 20/10 MIT problem set0/6 Practice∞

01Permutation Matrices and PA = LU 02Why RᵀR Is Always Symmetric 03Vector Spaces and Subspaces 04The Column Space of a Matrix Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

Vector Spaces & Subspaces

01Vector Spaces and Subspaces 02Column Space of a Matrix 03Null Space of a Matrix Problem set0/10 Problem set 20/10 Practice∞

01Rank, Pivots, and Free Variables 02Special Solutions of the Null Space 03Reading the Null Space from RREF Problem set0/10 Problem set 20/10 MIT problem set0/3 Practice∞

01Solvability of Ax = b 02Particular Solutions and the Null Space 03Rank and the Four Cases of Ax = b Problem set0/10 Problem set 20/10 MIT problem set0/8 Practice∞

01Linear Independence & the Null Space 02Basis and Dimension 03The Rank-Nullity Theorem Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01Rank and the Four Fundamental Subspaces 02Four Bases from One Row Reduction 03Matrices as Vector Spaces Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

01The Dimension Formula for Subspaces 02Differential Equations as Linear Algebra 03Rank-One Matrices and Outer Products 04The Four Fundamental Subspaces Problem set0/10 Problem set 20/10 MIT problem set0/1 Practice∞

01The Incidence Matrix as Difference Operator 02The Incidence Matrix of a Graph 03The Equilibrium Equation AᵀCAx = f Problem set0/10 Problem set 20/10 MIT problem set0/1 Practice∞

Orthogonality & Least Squares

01Orthogonal Subspaces 02Null Space as Orthogonal Complement 03The Normal Equations Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

01The Projection Matrix P = aaᵀ/aᵀa 02Projection onto Subspaces 03Projection and Least Squares Problem set0/10 Problem set 20/10 MIT problem set0/6 Practice∞

01Complementary Projections: P and I−P 02Least Squares and the Normal Equations 03Least Squares as Projection 04Invertibility of AᵀA Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01Orthogonal Matrices 02Projections onto Orthonormal Bases 03The Gram-Schmidt Process 04QR Factorization from Gram-Schmidt Problem set0/10 Problem set 20/10 MIT problem set0/3 Practice∞

Determinants

01The Three Axioms of the Determinant 02Determinant Properties from Three Axioms 03Determinants via Elimination 04Multiplicative and Transpose Properties Problem set0/10 Problem set 20/10 MIT problem set0/2 Practice∞

01The Permutation Formula for Determinants 02Cofactor Expansion 03Periodic Determinants of Tridiagonal Matrices Problem set0/10 Problem set 20/10 MIT problem set0/3 Practice∞

01The Cofactor Formula for Matrix Inverses 02Cramer's Rule 03Determinants as Volume Problem set0/10 Problem set 20/10 MIT problem set0/4 Practice∞

Eigenvalues & Dynamics

01Eigenvalues and Eigenvectors 02Finding Eigenvalues and Eigenvectors 03Complex Eigenvalues and Defective Matrices Problem set0/10 Problem set 20/10 MIT problem set0/2 Practice∞

01Diagonalization: A = SΛS⁻¹ 02Matrix Powers and the Stability Theorem 03Defective Matrices and Multiplicity 04Eigenvalues and the Fibonacci Sequence Problem set0/10 Problem set 20/10 MIT problem set0/3 Practice∞

01Solving du/dt = Au with Eigenvalues 02Eigenvalue Stability in the Complex Plane 03The Matrix Exponential e^(At)04The Companion Matrix Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01Why 1 Is Always an Eigenvalue of Markov Matrices 02Markov Chain Steady States via Eigenvalues 03Markov Matrices and Steady States 04Orthonormal Bases and Fourier Coefficients Problem set0/10 Problem set 20/10 MIT problem set0/7 Practice∞

Symmetric Matrices & the SVD

01The Spectral Theorem 02Why Symmetric Matrices Have Real Eigenvalues 03Sylvester's Law of Inertia 04Testing Positive Definiteness Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01The Conjugate Transpose 02Inverse of the Fourier Matrix 03The FFT Factorization Problem set0/10 Problem set 20/10 Practice∞

01Positive Definite Matrices 02Positive Definiteness via Pivots 03The Hessian and Positive Definiteness 04Principal Axis Theorem and the Ellipsoid Problem set0/10 Problem set 20/10 MIT problem set0/6 Practice∞

01Closure Properties of Positive Definite Matrices 02Matrix Similarity and Invariant Eigenvalues 03Why Eigenvalues Don't Classify Similarity 04Jordan Canonical Form Problem set0/10 Problem set 20/10 MIT problem set0/5 Practice∞

01The Singular Value Decomposition 02Reducing SVD to A^T A 03Computing the SVD: Two Worked Examples 04SVD and the Four Fundamental Subspaces Problem set0/10 Problem set 20/10 MIT problem set0/3 Practice∞

Linear Transformations & Applications

01Defining Linear Transformations 02Basis, Coordinates, and Linear Maps 03How a Linear Map Becomes a Matrix Problem set0/10 Problem set 20/10 Practice∞

01Image Compression as a Change of Basis 02Choosing a Basis: JPEG, Fourier, and Wavelets 03Change of Basis and Image Compression 04Change of Basis and Diagonalization Problem set0/10 Problem set 20/10 Practice∞

01Matrix Inverses and the Four Subspaces 02One-Sided Inverses and Projections 03The Pseudo-Inverse 04The Pseudo-Inverse via the SVD Problem set0/10 Problem set 20/10 Practice∞

Least Squares and the Normal Equations

When no exact solution exists for an overdetermined system, the normal equations A^TA x-hat = A^Tb minimize the sum of squared errors.

Loading…

←Previous Complementary Projections: P and I−P Next Least Squares as Projection →

Your summary note

1. Overdetermined fit and the squared error
  Turn the points $(1,1)$ , $(2,2)$ , $(3,2)$ and the line $y = C + Dt$ into three equations, write $A$ , $x$ and $b$ , and name the quantity $\|Ax-b\|^2$ being minimised.
2. The normal equations $A^TA\hat{x} = A^Tb$
  Compute $A^TA$ and $A^Tb$ for this example, solve the resulting two equations by elimination and back substitution, and carry it through to $y = \frac{2}{3} + \frac{1}{2}t$ .
3. The calculus route to the same system
  Write $E = (C+D-1)^2 + (C+2D-2)^2 + (C+3D-2)^2$ , set $\partial E/\partial C = 0$ and $\partial E/\partial D = 0$ , and match the two equations against the ones above.
4. Sensitivity to outliers
  Record the effect one far-off data point has on the fitted line once its error is squared, name such points outliers, and note that alternative methods exist.
Attempt 1 of 2