CPSC 491 Xin Liu Nov 17, 2010 Introduction • Xin Liu • PhD student of Dr. Rokne • Contact • liuxin@ucalgary.ca • Slides downloadable at • pages.cpsc.ucalgary.ca/~liuxin • The way to math world • Lecture attendance • Hard to learn by yourselves • Practices, practices, and practices … 2 Matrix-Vector Multiplication • Linear (the 1st degree) systems are the simplest, but most widely used systems in science and engineering • A basic problem: solving the linear equation system • Straight forward method • Gaussian elimination • Hard to do because • large scale • poor conditioned • small disturbance in coefficients causes big difference in solutions • A better method • SVD – Singular Vale Decomposition • Will be introduced gradually in a series of lectures 3 Definitions • An n-vector is defined as • Think about 3-vectors in Euclidean space • An mxn matrix is defined as • Multiplication 4 Linear Mapping • is a linear mapping, which satisfies • • Distributive law Associative law (for scalar) • Conversely, every linear map from Rn to Rm can be expressed as a multiplication by an mxn matrix • 5 Mat-vect multiplication • View the matrix-vector multiplication from another angle • If we write A as a combination of column vectors • Then the mat-vect multiplication can be written as • That means: b is a linear combination of the columns of A • 6 Mat-mat multiplication • Matrix-matrix multiplication • is defined as • We can calculate B columnwisely • Each column of B is a linear combination of the columns aj with the coefficients ckj • 7 Range • Definition: • The range of a matrix A, is the set of vectors that can be expressed as Ax for some x. • Theorem • range (A) is the space spanned by the columns of A. • The range of A is also called the column space of A. 8 Nullspace • Definition: • The nullspace (solution space) of A is the set of vectors x that satisfy Ax = 0. • Each vector x in the nullspace gives the expansion coefficients of the zero vector as a linear combination of columns of A 9 Rank • Column rank = dimension of space spanned by the matrix’s columns = # of linearly independent columns • Row rank = dimension of space spanned by the matrix’s rows = # of linearly independent rows • Row rank = Column rank = Matrix rank • Full rank • Theorem • 10 Inverse • A nonsingular or invertible matrix must be square and full rank. • The m columns of a nonsingular mxm matrix A span (form a basis) for the whole space Rm • • Any vector in Rm can be expressed as a linear combination of the columns of A The inverse of A is a matrix A-1, such that • • AA-1 = A-1A = I I is the mxm identity matrix • The inverse of a nonsingular matrix is unique. • A-1b is the unique solution of Ax = b. • A-1b is the vector of coefficients of the expansion of b in the basis of the columns of A. • 11 Transpose • Definition • The transpose AT of an mxn matrix A is nxm where the (i,j) entry of AT is the (j, i) entry of A. • Example • A is symmetric if A = AT. • Multiplication • • 12 Inner product • Inner product • Euclidean length • Angle • 13 Orthogonal vectors • Orthogonal (perpendicular) vectors • Vectors x, y are orthogonal if xTy = 0. • Orthogonal vector set • Orthogonal two vector sets • 14 Orthonormal • Definition • Theorem • Corollary • 15 Components of a vector • Inner products can be used to decompose arbitrary vectors into orthogonal components (project onto orthonormal vectors). • 16 Components of a vector 17 Orthogonal matrices • Definition: • According to the definition • Or • 18 An example • 2D rotation matrix 19 Multiplication by an orthogonal matrix • inner products is preserved • angles between vectors are preserved • lengths are preserved • 20