MS-A000* Matrix algebra ( Ville Turunen, Aalto University)

Size: px
Start display at page:

Download "MS-A000* Matrix algebra ( Ville Turunen, Aalto University)"

Transcription

1 x a c b Figure 1: Geometry of Pythagoras equation a 2 + b 2 = c 2, with x = a b MS-A000* Matrix algebra ( Ville Turunen, Aalto University) Welcome to the lecture course on linear algebra! We shall treat vectors and matrices in Euclidean spaces, the spectral theory (eigenvalue decompositions) of normal matrices, finally reaching the so-called SVD (Singular Value Decomposition) Such vector issues will be found everywhere in advanced mathematics and its applications in engineering and sciences Feedback is welcome: please contact villeturunen@aaltofi 0 It s all about geometry and algebra Vectors are best understood when illustrating the algebraic manipulations with schematic pictures For the best learning outcomes, the reader of these notes should draw his/her own mathematical pictures next to the calculations For example, see the picture visualizing Pythagoras equation There, by the areas of right-angled triangles and squares, c 2 = a b 2 + 4ab/2 = (a 2 + b 2 2ab) + 2ab = a 2 + b 2 Thus we get Pythagoras equation a 2 + b 2 = c 2 (1) Geometry of vectors is based on this! Actually, we may say without exaggeration that high-dimensional (even infinite-dimensional) calculations can often be correctly illustrated by just two-dimensional pictures! 1

2 (0, 0) x = (x 1, x 2 ) r = x ϕ = arg(x) Figure 2: Polar coordinates (r, ϕ) for point x = (x 1, x 2 ) R 2 1 Real plane and complex numbers Definition Let R be the real line Real plane R 2 consists of points x = (x 1, x 2 ), where x 1, x 2 R are the (Cartesian) coordinates of x R 2 Definition The polar coordinates (r, ϕ) = ( x, arg(x)) of point x = (x 1, x 2 ) R 2 satisfy r := x x 2 2; tan(ϕ) = x 2 x 1 if x 1 0 Here r 0 is distance of x from origin, and the argument ϕ = arg(x) R is the angle between positive horizontal axis and interval from O to x Remark: Example the argument ϕ is identified with ϕ + 2π here If x = ( 3, 1), then x = (2 cos(π/6), 2 sin(π/6)), since 3 2 x = + 12 = 2, arg(x) = arctan(1/ 3) = π/6 2

3 ir y x + y ir x x 0 R 0 x y x x x R Complex numbers Definition Figure 3: Arithmetic with complex numbers Points x = (x 1, x 2 ) R 2 can be thought as complex numbers x = x 1 + ix 2 C, where Re(x) := x 1 R is the real part and Im(x) := x 2 R is the imaginary part, and i := 0 + i1 is the imaginary unit, and t R is identified with t + i0 C If x, y C, there are familiar formulas: x + y := (x 1 + y 1 ) + i(x 2 + y 2 ), x y := (x 1 y 1 ) + i(x 2 y 2 ), x := ( x 1 ) + i( x 2 ) Complex conjugate of x = x 1 + ix 2 C is x = x := x 1 ix 2 C Absolute value of x C is (thinking of Pythagoras) x := x x 2 2 Example If x = 3 + i and y = 1 + 2i then x + y = 4 + 3i, x y = 2 i, x = 3 i, x = = 10 3

4 ir x + y y y 0 x x x + y x y R Figure 4: Triangle inequality of complex numbers Triangle inequality in C For x, y C, there is the triangle inequality Let s verify this First, calculating ( x + y ) 2 x + y 2 x + y x + y (2) = ( x + y ) 2 (x 1 + y 1 ) + i(y 1 + y 2 ) 2 = ( x x y + y 2) ( (x 1 + y 1 ) 2 + (x 2 + y 2 ) 2) = 2 x y 2 (x 1 y 1 + x 2 y 2 ), we need to show that x y x 1 y 1 + x 2 y 2 Here x 2 y 2 (x 1 y 1 + x 2 y 2 ) 2 = (x x 2 2)(y y 2 2) (x 2 1 y x 1 x 2 y 1 y 2 + x 2 2 y 2 2) = x 2 1 y x 2 2 y x 1 x 2 y 1 y 2 = (x 1 y 2 x 2 y 1 ) 2 0 This proved the triangle inequality (2) 4

5 1 i i 0 ir xy y x 1 R Figure 5: Product of complex numbers Product of complex numbers Product xy C of x, y C has the natural polar coordinate definition: { xy = x y, arg(xy) = arg(x) + arg(y) (3) Equivalently, in the Cartesian coordinates xy := (x 1 y 1 x 2 y 2 ) + i(x 1 y 2 + x 2 y 1 ) C (4) Especially, i 2 = (0 + 1i)(0 + 1i) = (0 1) + i(0 + 0) = 1 Equivalently, i 2 = i i = 1, and arg(i 2 ) = 2arg(i) = 2π/2 = π Example If x = i, y = + 5 i, then xy = (48 15) + i( ) 65 By the way, here x = y = 1 = xy = i Remark Always x x = (x 1 ix 2 )(x 1 + ix 2 ) = x x 2 2 = x 2 0 5

6 ir yx x + y y x y x + y y x yy xx (x + y)(x + y) 0 x R y x (x + y) xy Figure 6: Triangle inequality of complex numbers, again Another proof of Triangle inequality x + y 2 = (x + y)(x + y) = (x + y)(x + y ) = xx + yy + xy + yx = x 2 + y Re(xy ) x 2 + y xy = x 2 + y x y = ( x + y ) 2 Thus the Triangle inequality follows by taking square roots: x + y x + y 6

7 1 = e iπ i = e iπ/2 i 0 ir e it = cos(t) + i sin(t) t R 1 = e 0 = e i2π Figure 7: Euler s formula Division and powers of complex numbers Division x y = x/y C of x, y C (where y 0) satisfies { x/y = x / y, arg(x/y) = arg(x) arg(y) (5) Notice that x y = xy yy = xy y 2 Example If x = i and y = + 5 i then x y = xy y 2 y =1 = xy = ( ) + i( ) 65 = i Example If n Z then z n = z n and arg(z n ) = n arg(z) In case z = cos(t) + i sin(t), we get de Moivre s formula: (cos(t) + i sin(t)) n = cos(nt) + i sin(nt) (6) It can also be shown that the following Euler s formula e it = cos(t) + i sin(t) (7) holds, where the complex exponential is defined by the power series e z := k=0 1 k! zk (8) This satisfies for instance e w+z = e w e z 7

8 u v v u u v 1 v 2 O u + v v u 1 u 3 u v Figure 8: Schematic picture of vector operations 2 Vector spaces Vectors x = (x 1, x 2, x 3,, x n ) are everywhere : numbers x k are typically results of systematic measurements related to various phenomena To understand real vectors, we need complex numbers (see Gauss Fundamental Theorem of Algebra, on page 40; also Fourier transform easily takes real vectors to complex) From now on, let K := R or K := C (real or complex number field) Definition Vector space K n consists of points x = (x 1,, x n ), where x k K is the kth (Cartesian) coordinate of x K n Point O := (0,, 0) K n is the origin If λ K and x, y K n, let x + y := (x 1 + y 1,, x n + y n ), x y := (x 1 y 1,, x n y n ), λx := (λx 1,, λx n ), x := ( x 1,, x n ) Let a, b, x, y K n ; we identify vectors ab and xy if b a = y x Remark! We identify vector xy with point y x K n How do the vector operations look like? Illustrate these examples: Example Points λx, when x = (3, 1), π λ π Example λx + µy, when λ, µ { 1, 0, 1}, x = (3, 1), y = (1, 2) 8

9 u u O u v v v Geometry of vectors Figure 9: Norms, distances Definition The inner product (or dot product) of u, v C n is u v = u, v := n u k v k = u 1 v u n v n C (9) k=1 The norm u 0 of u C n is ( n ) 1/2 u := u, u 1/2 = u k 2 = ( u u n 2 ) 1/2 (10) k=1 u 2 = u, u can be thought as the energy of vector u C n Here, the number u k 0 is the distance from the equilibrium for each index k {1,, n}, with the energy proportional to the number u k 2 Then the sum of such energies is the total energy u 2 Definition Define the distance between u, v C n to be u v 0 (11) Remark Picture a triangle with vertices at O, u, v C n ; notice that u v 2 = u 2 + v 2 2 Re u, v Definition Vectors u, v C n are orthogonal if u, v = 0 (and then u v 2 = u 2 + v 2 ; remember Pythagoras!) Idea: u, v = u v cos(α), with α the angle between u, v Now here u, v = 0 means cos(α) = 0 (when u O v), the case of the orthogonality! 9

10 u O u/ u v/ v v Figure 10: Unit normalizations u/ u and v/ v of vectors u O and v O Cauchy Schwarz inequality Remember: Energy u, u = u 2 0 Can we estimate u, v C somehow? Proposition (Cauchy Schwarz inequality) For every u, v C n, u, v u v (12) Proof We may assume u v > 0 (for otherwise the claim is trivial) Also, noticing that u u, v = u v u, v, v we may assume the unit normalizations u = 1 = v, and then just prove that u, v 1 (Why this is enough? Think!) Indeed, now u, v = n u k v k k=1 (2) n u k v k ( ) k=1 n u k 2 + v k 2 k=1 2 = u 2 + v 2 2 u =1= v = 1 Above, inequality ( ) holds because 0 ( u k v k ) 2 QED 10

11 v O v u u u + v u u + v v x x y x z y y z z x z x y + y z u + v u + v Figure 11: Triangle inequality of vectors Triangle inequality (Corollary to Cauchy Schwarz) Triangle inequality For all u, v C n, u + v u + v (13) Proof By an application of Cauchy Schwarz (12), u + v 2 = u + v, u + v = u, u + u, v + v, u + v, v = u Re u, v + v 2 (12) u u v + v 2 = ( u + v ) 2 QED Example Let x = London, y = P aris, z = Rome For u = x y and v = y z, the Triangle inequality (13) then says: x z x y + y z The distance from x to z is at most the distance from x via y to z 11

12 u v u v v u + v u O u u + v v For a parallelogram in C n, the sum of the squares of the diagonals = the sum of the squares of the edges Figure 12: Parallelogram identity: u + v 2 + u v 2 = 2 u v 2 Polarization and parallelogram identities The norm was defined by the inner product Actually, the inner product can be recovered from the norm: Exercise Prove the polarization identity 4 u, v = u + v 2 u v 2 + i u + iv 2 i u iv 2 (14) Especially, notice that here 4 Re u, v = u + v 2 u v 2, Im u, v = Re u, iv Exercise Prove the parallelogram identity in C n : u + v 2 + u v 2 = 2 ( u 2 + v 2) (15) Draw a parallelogram in a plane, with vertices at O, u, v, u + v R 2 : are u + v, u v, u, v connected to the distances between these vertices? 12

13 u v R 3 O α v R 3 u R 3 Figure 13: Cross product Cross product in R 3 The cross product u v R 3 of vectors u, v R 3 is u v := (u 2 v 3, u 3 v 1, u 1 v 2 ) (u 3 v 2, u 1 v 3, u 2 v 1 ) (16) Equivalent geometric definition for the cross product: 1 u, u v = 0 = v, u v ; 2 u v = u v sin(α), with α the angle between u, v 3 The vector triple (u, v, u v) has right-handed orientation Application 1 u v is the area of the parallelogram with vertices at O, u, v, u + v R 3 Application 2 u v, w is the volume of the polytope with vertices at O, u, v, w, u + v, u + w, v + w, u + v + w R 3 Application 3 (q p) (r p) is a normal vector to a plane S R 3 if p, q, r S form a non-degenerate triangle 13

14 Nice to know: vector subspaces, dimension Definition (1) O V, Subset V K n is called a vector subspace if (2) x + y V whenever x, y V, and (3) λx V whenever λ K and x V The span of vectors S 1,, S k K n is the vector subspace { k } Z k = span{s j } k j=1 := λ j S j K n : λ 1,, λ k K For example, (this is a line or the origin), and j=1 Z 1 = KS 1 = {λ 1 S 1 : λ K} K n {O} Z 1 Z 2 Z k 1 Z k K n The dimensions are dim(k n ) = n and dim({o}) = 0; if here Z j+1 Z j, then Hence dim(z j+1 ) = 1 + dim(z j ) 0 dim(z k ) k Vectors S 1,, S k are linearly independent if dim(z k ) = k (otherwise they are called linearly dependent) Note The dimension depends on the scalar field K: The complex number field C has dimension 1 as a complex vector space If C is identified as the real plane R 2, then it has dimension 2 as a real vector space That is, Similarly, dim(c) = 1 < 2 = dim(r 2 ) dim(c n ) = n < 2n = dim(r 2n ), where we understand C n to be a complex vector space, even if sometimes it might be identified with the real vector space R 2n 14

15 Nice to know: Gram Schmidt orthonormalization Definition Vectors U 1,, U k K n are orthonormal if for all j, l {1,, k} { 1 when j = l, U j, U l = 0 when j l Gram Schmidt algorithm Let vectors S 1,, S k K n be linearly independent The Gram Schmidt process produces orthonormal vectors U 1,, U k K n as follows: Let U 1 := S 1 / S 1, (17) and for j 2 then where U j := Notice that for all d {1,, k} here V j V j, (18) j 1 V j := S j S j, U l U l (19) l=1 span{s 1,, S d } = span{u 1,, U d } Remark This Gram Schmidt process is numerically unstable (round-off errors accumulate), but there are ways to stabilize the process Example Vectors S 1 = and then V 2 = [ ] [ ] U 1 =, [ ] 3 and S 4 2 = [ ] 3 4 [ ] 3/5 4/5 [ ] 10 give 5 [ ] 1 3/ =, 2 4/5 [ ] 3/5 = 4/5 [ ] [ ] [ ] 3/5 44/5 =, 4/5 33/5 so that U 2 = V [ ] 2 4/5 V 2 = 3/5 15

16 A(u + v) = A(u) + A(v) A(v) K n : λu O u v u + v A K m : A(λu) = λ A(u) O = A(O) A(u) Figure 14: Idea of a linear mapping A : K n K m 3 Linear functions, linear equations Function A : K n K m is linear if A(u + v) = A(u) + A(v), A(λu) = λ A(u) for all λ K and u, v K n Then we often write Au := A(u) Equivalently, function A : K n K m is linear if and only if its graph G(A) is a vector subspace of K n K m = K n+m, where G(A) := {(u, A(u)) : u K n } Generic example Define function A : K n K m by n (A(u)) j := A jk u k, (20) where matrix elements A jk K for each j {1,, m} and k {1,, n} Then A is linear Actually, this example is surprisingly(?) general, as we shall soon see Trivial example We identify numbers x R with vectors (x) R 1 In this sense, function f : R R is linear if and only if f(x) = ax for some a R Then the graph G(f) = {(x, f(x)) : x R} described by the equation y = f(x) is a line through the origin (0, 0) of the (x, y)-plane, and this line has the slope a 16 k=1

17 j = (0, 1, 0) = I 2 j = (0, 1) = I 2 O i = (1, 0, 0) = I 1 O i = (1, 0) = I 1 k = (0, 0, 1) = I 3 Figure 15: Standard basis vectors of R 3 and R 2, respectively Norm The norm of linear A : K n K m is the number A := max Au = max Au, v (21) u K n : u 1 u K n,v K m : u, v 1 Then clearly Aw A w for all w K n Idea here: Linear functions treat vector operations nicely Norm measures the size of the mapping A (that is, how much A can stretch vectors in the extreme case) Standard basis vectors Definition Standard basis vectors I 1,, I n K n are defined by { 1 if j = k, I jk := 0 if j k For example, I 1, I 2, I 3 K 3 are I 1 = (1, 0, 0) = i, I 2 = (0, 1, 0) = j, I 3 = (0, 0, 1) = k Vectors u = (u 1,, u n ) K n can then be written u = n u k I k, k=1 which we shall use in the sequel Example: (9, 7, 5) = 9(1, 0, 0) + 7(0, 1, 0) + 5(0, 0, 1) 17

18 Matrix [A] of linear function A Let A : K n K m be linear Let I 1,, I n K n be the standard basis vectors If u = (u 1,, u n ) K n then Au K m, and Au u = A linear = n u k I k, k=1 n u k A(I k ), k=1 where A(I k ) K m Let A jk := (A(I k )) j K Then matrix A 11 A 12 A 1n [ ] A 21 A 22 A 2n A := Km n A m1 A m2 A mn (with m rows, n columns) has all the information about linear function as A : K n K m, (Au) j = In the sequel, we shall always identify n A jk u k (22) k=1 vector u = (u 1,, u n ) K n with matrix [u] = u 1 u n K n 1 We shall identify linear mapping A : K n K m with its matrix [A] K m n Notice that the kth column A k K m 1 of matrix [A] = [A 1 A n ] K m n is then A k = A 1k A 2k A mk = [A(I k)] K m 1, which is identified with the vector A(I k ) = (A 1k, A 2k,, A mk ) K m Especially, the standard basis vectors I 1,, I n K n for the so-called identity matrix [I] = [I 1 I n ] K n n, for which Iu = u for all u K n 18

19 A 1 = A(I 1 ) K n : O I n I 1 A K m : O = A(O) A n = A(I n ) Figure 16: Matrix [A] = [A 1 A n ] K m n of linear mapping A : K n K m Example Let A : R 3 R 2 be defined by A(u) := (2u 1 + 3u 2 + 4u 3, 5u 1 + 6u 2 + 7u 3 ) (23) Then A is linear, and its matrix [A] R 2 3 is [ ] [A] = The matrix notation for (23) would be [ ] [ ] [ ] [ ] Au = A u = u u u 3 [ ] 2u1 + 3u = 2 + 4u 3 5u 1 + 6u 2 + 7u 3 For instance, if u = ( 1, 2, 3) R 3, then [ ] [Au] = [A][u] = [ ] 2( 1) + 3( 2) + 4( 3) = = 5( 1) + 6( 2) + 7( 3) meaning that Au = ( 20, 38) R 2 We shall calculate the norm A later [ ] 20, 38 19

20 Matrices can be read from inner products: and v K m, then If A : K n K m is linear Au, v = m (Au) j v j = j=1 m n A jk u k v j j=1 k=1 Thus if I = [I 1 I 2 I 3 ] denotes the identity matrix of any dimension, if u = I k K n and v = I j K m, then Au, v = A jk (24) So, if we know all the inner products Au, v K for all u K n and v K m, we know all the matrix elements A jk K Remark Next we see that already the inner products Au, u are enough to determine A K n n when u C n (but not when u R n ) 20

21 A 2 = [ ] sin(ϕ) cos(ϕ) I 1 [ ] cos(ϕ) sin(ϕ) I 2 A 1 = [ cos(ϕ) sin(ϕ) O ϕ A = [A 1 A 2 ] = sin(ϕ) cos(ϕ) Identity matrix I = [I 1 I 2 ] = ] [ ] Figure 17: Rotation A R 2 2 by angle ϕ around the origin Weak formulation of linear mappings Weak Formulation Theorem Let A, B C n n Then A = B if Au, u = Bu, u for all u C n (25) Proof Let u, v, w C n, λ C, and Cw = (A B)w = Aw Bw Here Cw, w = (A B)w, w = Aw Bw, w = Aw, w Bw, w = 0 so that 0 0= Cw,w = λ C(u + λv), u + λv 0= Cw,w = λ 2 Cu, v + λ 2 Cv, u Plug in λ {1, i} to get the following pair of equations: { 0 = Cu, v + Cv, u, 0 = Cu, v Cv, u Clearly, Cu, v = 0 for all u, v C n So Aw Bw = Cw = O for all w C n Thus A = B QED Remark Complex numbers are needed in (25) If we only had Au, u = Bu, u for all u R n, we still could have A B For example, [ ] cos(ϕ) sin(ϕ) A = sin(ϕ) cos(ϕ) rotates R 2 by angle ϕ R around the origin Then Au, u = cos(ϕ) u 2, which does not identify A when cos(ϕ) < 1 21

22 K n : O x z y = tx + (1 t)z A K m : b = Ax = Ay = Az A(O) = O Figure 18: Linear equation Ax = b may have infinitely many solutions: If Ax = b = Az and x z then all the points y in the line passing through x and z are also solutions, as A(tx + (1 t)z) = = b for all t K 4 Linear equations, Gauss elimination For b K m and linear A : K n K m, the linear equation can have 0 or 1 or infinitely many solutions x K n Ax = b, (26) Remark: While the real-world problems are typically non-linear, they often can be locally approximated by linear equations with a good accuracy Matrix formulation Linear equation Ax = b can be written by matrices: that is equivalently [A]x = b, (27) A 11 A 12 A 1n x 1 b 1 A 21 A 22 A 2n x 2 = b 2, (28) A m1 A m2 A mn x n b m A 11 x 1 + A 12 x A 1n x n = b 1, A 21 x 1 + A 22 x A 2n x n = b 2, (29) A m1 x 1 + A m2 x A mn x n = b m 22

23 K n : x A 2 x = b 2 A K m : b = Ax O A 1 x = b 1 A(O) = O Figure 19: Linear equation Ax = b, with solutions x at the intersection of hypersurfaces A 1 x = b 1 and A 2 x = b 2 Geometric interpretation In Problem (29), the equation A j1 x 1 + A j2 x A jn x n = b j can be written as the hypersurface equation A j x = b j in K n, where A j = (A jk ) n k=1 = (A j1, A j2,, A jm ) K n In other words, such a hypersurface is the subset S j K n, where S j := {x K n : A j x = b j } That is, the problem Ax = b has solutions in the intersection S K n of hypersurfaces S 1, S 1,, S m, ie S := {x K n : A j x = b j for all j {1,, m}} So the linear problem has either 0 or 1 or infinitely many solutions: (0) zero solutions if S is empty, (1) one solution x if S = {x} (S has just one point x), ( ) infinitely many solutions if S contains at least two different points x and z (and then S contains the infinite line passing through x and z) Remark: by Without harm, the linear problem Ax = b can be abbreviated A 11 A 12 A 1n b 1 [ ] A 21 A 22 A 2n b 2 A b, that is A m1 A m2 A mn b m 23

24 K n : O x = (3, 1) (2, 5) x = 1 (4, 3) x = 9 Figure 20: In the example here, the linear equation [A][x] = [b] has a unique solution x = (3, 1) at the intersection of hyperspaces (2, 5) x = 1 and (4, 3) x = 9 (which are lines in the plane, in this case) Gauss eliminations Example [Clumsy version on the left; matrix version on the right:] { [ ] 2x 1 + 5x 2 = 1, or 4x 1 + 3x 2 = { [ ] 2x 1 + 5x 2 = 1, or 7x 2 = { [ ] 2x 1 = 6, or 7x 2 = { [ ] x 1 = 3, or x 2 = So the solution is x = (x 1, x 2 ) = (3, 1) (Check!) Geometric interpretation: this is the intersection of lines! 24

25 Sometimes no solutions Example 2x 1 + 3x 2 + 4x 3 = 2, 4x 1 + 3x 2 + 2x 3 = 6, 6x 1 + 6x 2 + 6x 3 = 9 2x 1 + 3x 2 + 4x 3 = 2, 3x 2 6x 3 = 2, 3x 2 6x 3 = 3 2x 1 + 3x 2 + 4x 3 = 2, 3x 2 6x 3 = 2, 0 = 1 or or or ; equation 0 = 1 is false; more precisely, the last row in the last matrix reads 0 x x x 3 = 1, which is a contradiction! So the original problem x x 2 = x 3 9 has no solution! Geometric interpretation: the original three (hyper)planes had empty intersection! In this example, each pair of the original (hyper)planes has an infinite line intersection (check this!) 25

26 Sometimes infinitely many solutions Example Ethanol C 2 H 5 OH burns (combusts with oxygen O 2 ), producing carbon dioxide CO 2 and water vapor H 2 O: the reaction is (x 1 ) C 2 H 5 OH + (x 2 ) O 2 (x 3 ) CO 2 + (x 4 ) H 2 O, where x 1, x 2, x 3, x 4 Z + Conservation of atoms C, H, O: 2x 1 1x 3 = 0, x 1 2x 4 = 0, or x 1 + 2x 2 2x 3 1x 4 = x 1 1x 3 = 0, x 3 2x 4 = 0, x 2 3x 2 3 1x 4 = / x 1 1x 3 = 0, x x 3 1x 4 = 0, 0 2 3/ x 3 2x 4 = x 1 2x 3 4 = 0, x 2 2x 4 = 0, x 3 2x 4 = We obtain 1x 1 1x 3 4 = 0, +1x 2 1x 4 = 0, +1x 3 2x 3 4 = 0 When x 4 = 3t R, we get or x = (x 1, x 2, x 3, x 4 ) = (t, 3t, 2t, 3t) R 4, where t R (infinitely many solutions!) The smallest x j Z + are so that the reaction formula is x = (1, 3, 2, 3), C 2 H 5 OH + 3 O 2 2 CO H 2 O (Check!) Geometric interpretation: The intersection of the original hypersurfaces in R 4 is the line { (t, 3t, 2t, 3t) R 4 : t R } 26

27 Gauss elimination process in nutshell What happens in the Gauss elimination? Write linear equation Ax = b as the matrix [ A b ], on which we apply row operations (here row = equation): 1 add weighted row to another row; 2 exchange order of two rows; 3 multiply a row by a constant λ 0; If we get from Ax = b another linear problem Cx = d, we denote [ A b ] [ C d ] Our goal is a matrix [C], where 1 next row should not have less initial zeros than the row above; 2 each column has at most one non-zero entry µ (and you may still normalize the row with division by µ) Write solution(s) in form (or show that there are no solutions!) x = (x 1,, x n ) = Remark! When finding solution candidates x, it is best to check that really Ax = b 27

28 K n : u A K m : Au B K p : BAu O AO = O BO = O Figure 21: Composing linear mappings A : K n K m and B : K m K p gives a linear mapping BA := B A : K m K p 5 Matrix product [B][A] = [BA], matrix inversion [A] 1 = [A 1 ] Let A, B be linear such that K n A K m B K p Then K n BA K p is linear, and we define matrix product [B] [A] := [BA] K p n Now because (BAu) i = m B ik (Au) k = m (BA) ij = B ik A kj, (30) m B ik k=1 ( n n m ) A kj u j = B ik A kj u j k=1 k=1 j=1 j=1 k=1 Remark Let A, B be linear, where { A : K n 1 K m 1, B : K n 2 K m 2 For BA to be defined, we must have m 1 = n 2 Thus [B][A] := [BA] is defined only if the number of columns in [B] is the number of rows in [A] Then we have the norm estimate because BAu B Au B A u BA B A, (31) 28

29 Example Let 9 8 [ ] [B] = and [A] = Now A : R 2 R 2 and B : R 2 R 3, so A + B, AB and BB cannot be defined Similarly, we cannot compute [A] + [B] = [A + B], we cannot compute [A][B] = [AB], we cannot compute [B][B] = [BB]! But BA : R 2 R 3 and A 2 = AA : R 2 R 2 can be defined, and 9( 1) + 8( 3) 9( 2) + 8( 4) 33 [B][A] = 7( 1) + 6( 3) 7( 2) + 6( 4) = 25 5( 1) + 4( 3) 5( 2) + 4( 4) , 26 [A] 2 := [A][A] = = = [ ] [ ] [ ] ( 1)( 1) + ( 2)( 3) ( 1)( 2) + ( 2)( 4) ( 3)( 1) + ( 4)( 3) ( 3)( 2) + ( 4)( 4) [ ]

30 R : f S : f(x) x y f(y) Figure 22: Injective function f : R S does not destroy information! If x y for x, y R and f : R S injective then f(x) f(y) Properties of functions Informally, function f : R S between sets R, S is a rule that on each input x R produces an output f(x) S The f-image f(q) S of a subset Q R is Function f : R S is f(q) = {f(x) : x Q} injective provided that f(x) = f(y) only if x = y R; surjective if f(r) = S (that is, the f-image covers S completely); bijective if it is both injective and surjective In this case, f : R S has the inverse function f 1 : S R, where f(x) = u x = f 1 (u) What if here R, S are vector spaces and f : R S is linear? 30

31 K n : x = A 1 (u) y = A 1 (v) A K m : u = Ax v = Ay O = A 1 (O) B=A 1 O = AO Figure 23: Inverse function B = A 1 of a linear bijection A : K n K m is automatically linear, and then also m = n Bijective linear functions Theorem Let A : K n K m be a linear bijection Then A 1 : K m K n is linear, and m = n Proof Let us denote B = A 1, and let u, v K m, λ K Then u = A(x) and v = A(y) for some x, y K n, and linearity of B follows: B(u + v) = B(A(x) + A(y)) = B(A(x + y)) = x + y = B(u) + B(v), B(λu) = B(λ A(x)) = B(A(λx)) = λx = λ B(u) Notice that the origin x = O K n is the unique solution to A(x) = O K m ; the solution could not be unique if m < n (here A(x) = O is a system of m equations with n unknown variables x 1,, x n K) Hence m n But symmetrically n m: here u = O K m is the unique solution to A 1 (u) = O K n 31 QED

32 Inverse matrix [A] 1 = [A 1 ] Let I = (x x) : K n K n be the identity function If A, B, C : K n K n are linear and BA = I = AC then B = C = A 1, because B = BI = B(AC) = (BA)C = IC = C Let A : K n K n be a linear bijection; we know that then A 1 : K n K n is also a linear bijection Then the inverse matrix of A is defined to be the matrix of A 1, that is [A] 1 := [A 1 ], and we say that [A] is invertible: here A 1 A 1, because 1 = I = A 1 A A 1 A Finding X = A 1 means solving a linear equation AX = I, which can be done by the Gauss elimination [A I] [I X] = [I A 1 ] Example [I] 1 = [I 1 ] = [I], that is: [ ] [ ] 1 [ ] = 1, = 0 1 [ ] 1 0, = 0 1 0, Example = [ ] [ ] Example If A = and B = then B = A : [ ] [ ] [ ] [ ] (3) + 1( 5) 2( 1) + 1(2) 1 0 = = = I, (3) + 3( 5) 5( 1) + 3(2) 0 1 [ ] [ ] [ ] [ ] (2) 1(5) 3(1) 1(3) 1 0 = = = I (2) + 2(5) 5(1) + 2(3)

33 [ ] 2 1 Example Let [A] = Then 5 3 [ ] [ ] A I = [ ] [ ] [ ] [ ] 3 1 Thus [A] 1 = [X] = 5 2 = [ I X ] Remark! It is good to check that [X] is the inverse matrix to [A]: [A][X] = = [I], [X][A] = = [I] (It is enough to check only [A][X] or [X][A] why?) Example Let us try to find the inverse to [A] = : [ A ] I = The equations in the last row are now 0 = 1, 0 = 2, 0 = 1 which is clearly false! The reason is that here matrix [A] R 3 3 is not invertible, ie linear function A : R 3 R 3 is not bijective! 33

34 O I n Q I 1 A AO = O A 1 = A(I 1 ) A(Q) A n = A(I n ) Figure 24: Let A : R n R n be linear Let Q = [0, 1] n R n be the unit cube Then det[a] 0 is the volume of the polyhedron A(Q) R n, and det[a] < 0 would mean flipping the orientation 6 Determinant The determinant det[a] K of a matrix [A] K n n describes geometric information about the linear function A : K n K n If a subset S K n has n-dimensional volume 1 then: A(S) K n has n-dimensional volume det[a] When K = R: if det[a] < 0 then A(S) is a mirror image of S; if det[a] > 0 then A(S) and S have the same orientation We shall need four laws (L1,L2,L3,L4) to find determinants: (L1) det[i] = 1 for identity matrix [I] K n n (L2) If one column is multiplied by λ K, then det is multiplied by λ (L3) If matrix has two same columns then determinant is 0 (L4) Let X j be the jth column of X = [X 1 X n ] K n n If C k = A k + B k and C j = A j = B j whenever j k then det[c] = det[a] + det[b] Example For [A] = 0 3 0, laws (L1,L2) yield det[a] = det = det (L1) = = 30 (This matrix [A] here stretches the standard basis by the respective factors 2, 3, 5, thus strecthing the 3-dimensional volumes by the factor = 30) 34

35 [ ] 0 = I 1 2 [ ] 0 = O 0 Q I 1 = [ ] 1 = I I 2 [ ] 1 0 a A= A 1 = A(I 1 ) = b c d [ ] a c A(Q) O = AO A(I 1 + I 2 ) = A 1 + A 2 = A 2 = A(I 2 ) = [ ] b d [ ] a + b c + d Figure 25: Matrix [A] R 2 2 maps the unit cube (the usual unit square) Q = [0, 1] 2 = {x = (x 1, x 2 ) R 2 : x 1, x 2 [0, 1]} to the parallelogram A(Q) R 2, which has the 2-dimensional volume (or the usual area) det[a] 0 Law (L5) (Corollary to (L3,L4)) two columns! Sign of det changes by interchanging Proof: It is enough to check the 2-dimensional case: [ ] [ ] [ ] a + b a + b (L4) a a + b b a + b = det = det + det c + d c + d c c + d d c + d [ ] [ ] [ ] [ ] (L4) a a a b b a b b = det + det + det + det c c c d d c d d [ ] [ ] (L3) a b b a = +det + det QED c d d c 0 (L3) Example: [ ] a b det c d The general computation for the 2-dimensional case is [ ] [ ] (L4) a b 0 b = det + det 0 d c d [ ] [ ] [ ] [ ] (L4) a b a 0 0 b 0 0 = det + det + det + det d c 0 c d [ ] [ ] [ ] [ ] (L2) = ab det + ad det + bc det + cd det [ ] (L3,L5) 1 0 = (ad bc) det 0 1 (L1) = ad bc 35

36 More generally: Similarly for any [A] K n n, det[a] = a P det[p ], (32) [P ] permutation where the permutation matrices [P ] R n n have elements P jk {0, 1} with exactly one 1 in each row and in each column; a P K is the product of those A jk for which P jk = 1 For dimension n, there are exactly n! permutation matrices For instance, for n = 2 we have n! = 2 permutations, and [ ] [ ] a b a 0 det = det c d 0 d +det [ ] 0 b c 0 = ad det [ ] bc det [ ] = ad bc Example For n = 3, there are n! = 6 permutation matrices [P ] R 3 3, which are , 0 0 1, 1 0 0, 0 0 1, 1 0 0, Find the determinants of these permutation matrices! Notice that a b c det d e f = g h i [ a 0 0 ] [ a 0 0 ] [ 0 b 0 ] [ 0 b 0 ] [ 0 0 c ] [ 0 0 c ] = det 0 e 0 + det 0 0 f + det 0 0 f + det d det d det 0 e i 0 h 0 g i 0 h 0 g 0 0 = +aei afh + bfg bdi + cdh ceg Remark! By (32), Laws (L1,L2,L3,L4,L5) are ok when the word column is changed to the word row Thus if we obtain [B] K n n from [A] K n n by the Gauss elimination (not multiplying single rows) then det[a] = ( 1) k det[b], where k is the number of the row interchanging operations Example: det[a] = = 42, where [A] =

37 Remark! Clearly, A K n n is invertible if and only if it can be transformed to the identity matrix I K n n by the Gauss row operations: [A I] Gauss [I A 1 ] On the other hand, invertibility is equivalent to det[a] 0, since the determinant can be calculated by the Gauss row operations Sub-determinants For finding determinants, the Gauss row operations are most useful if we know the exact numerical values of the matrix elements For more general determinant formulas, application of (32) might be better Related to this, we may go from dimension n to dimension n 1 by subdeterminants: for [A] K n n, let [B (j,k) ] = [omit row j and column k from A] K (n 1) (n 1), det[a] = = n ( 1) j+k A jk det[b (j,k) ] j=1 n ( 1) j+k A jk det[b (j,k) ] k=1 Here the signs have the pattern [ ( 1) j+k] n j,k=1 = Example Let us first develop the sub-determinants with respect to the first row [ a b c ] b, and then with respect to the second column e : h a b c [ ] [ ] [ ] det d e f e f d f d e = +a det b det + c det h i g i g h g h i [ ] [ ] [ ] d f a c a c = b det + e det h det g i g i d f 37

38 Row operations as matrix products Any simple Gauss row operation B EB on matrix B K n n can be written as an elementary matrix product EB K n n : for instance, a b c a b c d e f = g h i g h i d e f corresponds to interchanging rows (2nd and 3rd), a b c a b c 0 λ 0 d e f = λd λe λf g h i g h i corresponds to multiplying a row (2nd) by constant λ K, and a b c a b c 0 1 λ d e f = d + λg e + λh f + λi g h i g h i corresponds to adding a λ-weighted row (3rd) to another row (2nd) Theorem Determinant is multiplicative: for all A, B K n n det[ab] = det[a] det[b] (33) Proof For non-invertible A this is trivial An invertible A is of the form A = E 1 E 2 E k, (34) where each E j K n n corresponds to a Gauss row operation as above Here for all M K n n Thereby Corollary: det[ab] det[e j M] = det[e j ] det[m] (35) (34) = det[e 1 E 2 E k B] (35) = det[e 1 ] det[e 2 E k B] (35) = det[e 1 ] det[e 2 ] det[e k ] det[b] (35) = det[e 1 E 2 E k ] det[b] (34) = det[a] det[b] QED If A K n n is invertible then because 1 = det[i] = det[aa 1 ] = det[a] det[a 1 ] det[a 1 ] = 1/det[A], (36) 38

39 z O x Cx A O = AO Az = λz Cx Ax = λx Figure 26: Eigenvectors x, z C n on the same eigenline Cx of A C n n 7 Eigenvalues and eigenvectors Linear A : C n C n (and corresponding matrix [A] C n n ) has eigenvector x C n (or [x] C n 1 ) and eigenvalue λ C if Ax = λx, where eigenvector x O (of course, trivially AO = O = λo, where O C n is the origin) Remark Notice that if t C then here A(tx) = ta(x) = λ(tx): thus it would be appropriate to talk about the eigenline Cx = {tx C n : t C}, where tx C n is another eigenvector whenever t 0 How to find all the eigenvalues? Notice that Ax = λx (A λi)x = O A λi linear, x O A λi not invertible, which means det[a λi] = 0, so you may find the eigenvalues λ C by solving this so-called characteristic equation How to find the eigenvectors? Let λ C be an eigenvalue of a square matrix A C n n The corresponding eigenvectors x C n satisfy Ax = λx (equivalently (A λi)x = O), and we find them by the Gauss elimination on [A λi O] Just remember that the origin O C n is never an eigenvector 39

40 Eigenvalues as roots of characteristic polynomial Definition The characteristic polynomial of [A] C n n is p A : C C, where p A (z) := det[a zi] By the Fundamental Theorem of Algebra [Gauss], polynomials split uniquely in C into product of first order terms; thus p A (z) = ( 1) n (z λ 1 ) (z λ n ), where λ 1,, λ n C are the eigenvalues of A The algebraic multiplicity d = m a (λ) of an eigenvalue λ C is the degree d of the factor (z λ) d in p A The geometric multiplicity m g (λ) of an eigenvalue λ C is the dimension of the vector subspace spanned by the corresponding eigenvectors Notice that always Actually, 1 m g (λ) m a (λ) n p A (z) = ( 1) n ( z n tr[a]z n ( 1) n det[a] ), where tr[a] C is the trace of [A] C n n, satisfying (typically here A kk λ k ) tr[a] := n A kk = k=1 n λ k (37) k=1 40

41 ir A λ 0 R Figure 27: λ A if λ C is an eigenvalue of A C n n Eigenvalues/matrix norm Recall that the norm of matrix A C n n is A := max u C n : u 1 Au = max Au, v u,v C n : u, v 1 If λ > A then λ C cannot be an eigenvalue of A: This follows from n (λi A) (A/λ) k k=0 = (λi A) ( I + (A/λ) 1 + (A/λ) (A/λ) n) = λii AI + λi(a/λ) 1 A(A/λ) 1 + λi(a/λ) 2 A(A/λ) n = λi λ(a/λ) n+1 n λi, where the limit exists, because (A/λ) n+1 = A n+1 / λ n+1 n+1 n ( A / λ ) 0 since here A / λ < 1 Hence λi A C n n is invertible, and (λi A) 1 = λ 1 k=0 (A/λ) k Moreover, by applying the geometric series, we obtain the norm estimate (λi A) 1 λ 1 A k / λ k = k=0 1 λ A 41

42 a 0 0 Example If [A] = 0 b 0 C 3 3 then 0 0 c p A (z) = det[a zi] a z 0 0 = det 0 b z c z giving eigenvalues a, b, c C Moreover, where = ( 1) 3 (z a)(z b)(z c), p A (z) = ( 1) 3 ( z 3 (a + b + c)z 2 + (ab + ac + bc)z abc ), tr[a] = a + b + c C, det[a] = abc C [ ] a b Example If [A] = C c d 2 2 then p A (z) = det[a zi] [ ] a z b = det c d z = (a z)(d z) bc = ( 1) 2 (z 2 (a + d)z + (ad bc)) = ( 1) 2 (z λ 1 )(z λ 2 ), where eigenvalues λ 1, λ 2 C can be found easily Notice that here tr[a] = a + d C, det[a] = ad bc C 42

43 Example Let R : R 2 R 2, for which R(x) = (x 2, x 1 ), that is [ ] 0 1 [R] = 1 0 Here det[r] = 1 R reflects the space with respect to the hyperplane (line) {x R 2 : x 1 = x 2 } Now p R (z) = det[r zi] [ ] 0 z 1 = det 1 0 z = z 2 1 = (z 1)(z + 1), giving eigenvalues λ = ±1 Corresponding eigenvectors x? R(x) = λx? We solve (R λi)(x) = O by the Gauss elimination: [ ] [ ] [ ] λ 1 0 Gauss 0 1 λ 2 0 λ=± [R λi O] = = 1 λ 0 1 λ 0 1 λ 0 Eigenvectors for λ = +1: x = (t, t) C 2 for 0 t C Eigenvectors for λ = 1: x = (t, t) C 2 for 0 t C Example Let Q : R 2 R 2, for which Q(x) = (x 2, x 1 ), that is [ ] 0 1 [Q] = 1 0 Here det[q] = +1 Q rotates the space around the origin Now p Q (z) = det[q zi] [ ] 0 z 1 = det 1 0 z = z = (z i)(z + i), giving eigenvalues λ = ±i Corresponding eigenvectors x? Q(x) = λx? We solve (Q λi)(x) = O by the Gauss elimination: [ ] [ ] [ ] λ 1 0 Gauss λ 2 0 λ=±i [Q λi O] = = 1 λ 0 1 λ 0 1 λ 0 Eigenvectors for λ = +i: x = (t, +it) C 2 for 0 t C Eigenvectors for λ = i: x = (t, it) C 2 for 0 t C 43

44 I n = (0, 0,, 1) O I 1 = (1, 0,, 0) Λ:C n C n O = ΛO λ 1 I 1 λ n I n Figure 28: Action of a diagonal matrix Λ C n n with Λ kk = λ k 8 Diagonalization Matrix Λ = [Λ jk ] n j,k=1 Cn n is diagonal if Λ jk = 0 whenever j k That is, with λ k := Λ kk C we have λ λ 2 0 Λ = λ n Let I 1, I 2,, I n R n 1 be the standard basis vectors Here obviously Λ(I k ) = λ k I k, meaning that I k is an eigenvector corresponding to the eigenvalue λ k C Diagonal matrices are easy to sum and to multiply, eg here (λ 1 ) Λ (λ 2 ) = (λ n ) 999 Also, here Λ is invertible if and only if λ 1 λ 2 λ 3 λ n 0, and then (λ 1 ) Λ 1 0 (λ 2 ) 1 0 = (λ n ) 1 All this is to say: diagonal matrices are really easy to handle It turns out that many other matrices can be diagonalized, making them easy to treat when seen from a suitable viewpoint 44

45 S n O S 1 A:C n C n A(S n ) = λ n S n O = AO A(S 1 ) = λ 1 S 1 S 1 = [S 1 S n ] 1 S = [S 1 S n ] I n = S 1 (S n ) O I 1 = S 1 (S 1 ) diagonal Λ:C n C n λ n I n O = ΛO λ 1 I 1 Figure 29: Diagonalization Λ = S 1 AS In other words, A = SΛS 1 Diagonalization Matrix A C n n can be diagonalized if there is invertible S C n n for which Λ = S 1 AS C n n is diagonal, ie Λ jk = 0 whenever j k: then A = SΛS 1, and (writing λ k := Λ kk C) we have λ λ 2 0 Λ =, 0 0 λ n where λ 1, λ 2,, λ n C are the eigenvalues of A The kth column S k of S = [S 1 S n ] is the eigenvector of A corresponding to the eigenvalue λ k C: (AS) jk = (SΛS 1 S) jk = (SΛ) jk = n S jl Λ lk = λ k S jk l=1 We may say that a diagonalizable matrix A = SΛS 1 looks like a diagonal matrix Λ after changing the coordinate point of view by S 45

46 Remark: Such diagonalization A = SΛS 1, if it exists, may not be unique, as it changes if the choose different eigenvectors, or if we permute the order of the eigenvalues (and the corresponding order of the eigenvectors) For instance, c [ a b d ] [ λ1 0 0 λ 2 c ] [ ] 1 a b = d [ t2 b t 1 a t 2 d t 1 c ] [ λ2 0 0 λ 1 ] [ t2 b t 1 a ] 1 t 2 d t 1 c whenever ad bc 0 t 1 t 2 Here, eigenvectors [ ] [ ] a b t 1, t c 2 d respectively correspond to the eigenvalues λ 1, λ 2 C Nevertheless, we have: Theorem Matrix A C n n can be diagonalized if and only if m g (λ) = m a (λ) for all the eigenvalues λ of A Remark: For a matrix A C n n, remember that 1 m g (λ) m a (λ) n always Matrix A C n n cannot be diagonalized if m g (λ) < m a (λ) for just one eigenvalue λ of A, as then we cannot find enough eigenvectors to achieve the invertible eigenvector matrix S eventually! In the last two examples in the previous Chapter (Eigenvalues and eigenvectors), the eigenvalues of the diagonalizable matrices R, Q : C 2 C 2 have all multiplicities 1 Example Let [A] = [ ] 1 b, where b 0 (so [A] is not diagonal) Now 0 1 p A (z) = det[a zi] [ ] 1 z b = det 0 1 z = (1 z) 2, giving algebraic multiplicity m a (λ) = 2 for eigenvalue λ = 1 As b 0, then (A λi)(x) = O has only solution x = (t, 0) for constants t C So the geometric multiplicity m g (λ) = 1 < m a (λ) This is an example of a non-diagonalizable matrix! In other words, here S 1 AS is not diagonal no matter which invertible S we choose 46

47 L U 11 U 12 U 13 U 1n L 21 L U 22 U 23 U 2n L = L 31 L 32 L 33 0, U = 0 0 U 33 U 3n L n1 L n2 L n3 L nn U nn Figure 30: Lower triangular L C n n, upper triangular U C n n Example Let λ C be the only eigenvalue of diagonalizable A C n n Thus A = S(λI)S 1 = λss 1 = λi That is, A = constant times identity Example Matrix M C n n is triangular if it is lower triangular (M jk = 0 whenever j < k) or upper triangular (M jk = 0 whenever j > k) If triangular A has zero diagonal, then λ = 0 is its only eigenvalue; by the previous example, such A can be diagonalized only if A = O Example AS = SΛ, if [ ] a b A =, S = 0 0 [ ] 1 b, Λ = 0 a [ ] a Thus if a 0 then we have A = SΛS 1 (if a = 0 then S is not invertible) Remark On page 57 we learn that matrix is normal (see page 50) if and only if it has a unitary diagonalization In the previous example above, A = SΛS 1 is a normal matrix if and only if b = 0: thus some non-normal matrices can be diagonalized (yet not unitarily diagonalized) 47

48 Functions of square matrices Analytic function f : C C can be presented as power series f(z) = c k z k (38) k=0 Eg functions exp, cos, sin are analytic Define f : C n n C n n by f(a) = c k A k (39) Here f(a) ij f(a ij ) often! But with diagonalization A = SΛS 1, ( ) f(a) = c k (SΛS 1 ) k = = S c k Λ k S 1 = S f(λ) S 1, k=0 where f(λ) C n n is diagonal with f(λ) jj = f(λ jj ) C Nice! Example Then where k=0 k=0 Let A = SΛS 1, where a 0 0 Λ = 0 b c exp(a) = S exp(λ) S 1, A 1/2 = S Λ 1/2 S 1, e a 0 0 a 0 0 exp(λ) = 0 e b 0, Λ 1/2 = 0 b e c 0 0 c So, what would be A then? If A C n n and smooth func- Application to differential equations: tions x 1,, x n : R C such that x (t) = A x(t) (where (x (t)) j = (x j ) (t) = d dt x j(t), naturally), then x(t) = exp(ta) x(0) 48

49 K n : O A w u, u = 1 A A K m : A(O) = O w, w = 1 Au Figure 31: Adjoint A : K m K n is the mirror image of linear A : K n K m in the sense that u, A w = Au, w for all u K n and w K m 9 Adjoint matrices For a linear mapping A : K n K m, the adjoint is the linear mapping A : K m K n defined by the inner product duality Au, w =: u, A w (40) for all u K n and w K m Why this definition works? Indeed, Au, w = m (Au) j w j = j=1 m n A jk u k w j = j=1 k=1 ( n m ) u k A jk w j, k=1 j=1 which shows that the adjoint A has the matrix elements (A ) kj = A jk K Clearly (A ) = A There is the norm equality A = A, because A = max Au, v u C n,v C m : u, v 1 = max u, u C n,v C m : u, v 1 A v = A Example [ ] 1 + 2i 3 + 4i 5 + 6i Let A = C 7 + 8i i i Its adjoint matrix is 1 2i 7 8i A = 3 4i 9 10i C i 11 12i 49

50 Important classes of matrices Definition Let us introduce the following classes of square matrices: Matrix P C n n is positive if P x, x 0 for all x C n Matrix S C n n is symmetric (or self-adjoint) if S = S Matrix N C n n is normal if N N = NN Matrix U C n n is unitary if U = U 1 : then U(x), U(y) = x, y for all x, y C n ; such U preserves distances and angles in C n! Remark These classes of matrices have some connections: Above, P ositive Symmetric, and Symmetric Normal, but Normal Symmetric, and Symmetric P ositive Also, Unitary Normal, but Normal Unitary Term positive is often more accurately positive semi-definite λ λ 2 0 Exercise Let us study a diagonal matrix Λ = Cn n 0 0 λ n For which λ k C the following properties hold? (a) Λ is positive (b) Λ is symmetric (c) Λ is normal (d) Λ is unitary Remark Notice that U U = I means that the columns U k C n 1 of unitary U = [U 1 U n ] C n n are mutually orthonormal to each other: U k, U k = U k 2 = 1, and U j, U k = 0 for j k Moreover, if A C n n is positive/symmetric/normal/unitary then also U AU has the same property Remark Invertibility of S = [S 1 S n ] C n n means that the vectors S 1,, S n are linearly independent; out of these vectors Gram Schmidt gives orthonormal U 1,, U n, and then U = [U 1 U n ] C n n is unitary Definition For A R m n, the adjoint is the transpose A T = A A real unitary matrix A R n n is called orthogonal: this means A T A = I = A A T, in other words A T = A 1 here 50

51 Eigenvalues of symmetric / positive matrices Claim: Symmetric matrix has real eigenvalues Proof Let λ C be an eigenvalue of a symmetric matrix S = S C n n with eigenvector x C n 1 Then λ x, x = λx, x Sx=λx = Sx, x = x, S x S =S = x, Sx Sx=λx = x, λx = λ x, x Thus λ = λ (ie λ R), because x, x = x 2 0 QED Claim: Positive P C n n is symmetric, having eigenvalues 0 Proof By the Weak Formulation Theorem we have P = P, as P x, x = x, P x real = x, P x = P x, x for any x C n Let P u = λu, where u O eigenvector with eigenvalue λ C Then 0 P u, u = λu, u = λ u, u = λ u 2, so that 0 λ QED 51

52 Q(K n ) = Z Qx O x = P x + Qx P x Z = P (K n ) Figure 32: Orthogonal projection P K n n onto vector subspace Z K n Then the linear mapping Q = I P is the orthogonal projection onto the orthogonal vector subspace Z = {u K n : u, z = 0 for all z Z} Nice to know: orthogonal projections For a vector subspace Z K n of dimension d = dim(z) there are linearly independent vectors S 1,, S d K n such that Z = span{s 1,, S d } From the sequence (S 1,, S d, I 1, I 2,, I n ), after throwing away the early vectors that violate the linear independence, the Gram Schmidt process finds the orthonormal vectors U 1,, U n K n such that Z = span{u 1,, U d } If we define P : K n K n by P x := d x, U k U k, (41) then P is the orthogonal projection onto Z, satisfying P = P = P 2 k=1 Remark: If P K n n satisfies P = P = P 2 then P is the orthogonal projection onto Z := P (K n ) The idea: P x Z is the closest point in Z to x K n In other words, P x is the perpendicular shadow of point x on subspace Z Now let Q = I P, that is x = P x + Qx (for all x K n ), then we have the Pythagorean identity x 2 = P x 2 + Qx 2 Here, Q is the orthogonal projection onto the orthogonal vector subspace Z = {u K n : u, z = 0 for all z Z} Notice that P Q = O = QP, and P (K n ) = Z = {x K n : Qz = O}, Q(K n ) = Z = {x K n : P z = O} P = [P 1 P n ] K n n projects orthogonally onto Z = P (K n ) = span{p j } n j=1 52

53 S n O S 1 A:C n C n A(S n ) = λ n S n O = AO A(S 1 ) = λ 1 S 1 S 1 = [S 1 S n ] 1 S = [S 1 S n ] I n = S 1 (S n ) O I 1 = S 1 (S 1 ) diagonal Λ:C n C n λ n I n O = ΛO λ 1 I 1 Figure 33: Diagonalization Λ = S 1 AS In other words, A = SΛS 1 10 Unitary diagonalization Remember the following matrix properties: M C n n is symmetric (or self-adjoint) if M = M B C n n is normal if B B = B B U C n n is unitary if U = U 1 Also, remember what the diagonalization Λ = S 1 AS for a matrix A C n n means: previously we just required S C n n to be invertible In unitary diagonalization, we want S unitary: remember that unitary matrices preserve inner products and norms! In order for this to be possible, something has to be assumed from A, but fortunately this assumption turns out to be quite easy and natural: namely, A should be normal! [ ] [ ] 0 b 0 c Example Let A =, where b, c C Then A = c 0 b Thus 0 here A is symmetric if and only if c = b Moreover, [ ] [ ] [ ] 0 c A 0 b c 2 0 A = b = 0 c 0 0 b 2, [ ] [ ] [ ] 0 b 0 c AA b 2 0 = c 0 b = 0 0 c 2 Hence here A is normal if and only if b = c 53

54 A 11 A 12 A 13 A 1n 0 A 22 A 23 A 2n A = 0 0 A 33 A 3n A nn A A=AA = A = A A A A nn Figure 34: Any normal (upper or lower) triangular matrix is diagonal! And what if the matrix would be symmetric and triangular? Normal triangular matrix is diagonal! Let A C n n Then (A A) kk = (AA ) kk = n A lk A lk = l=1 n A kl A kl = l=1 n A lk 2, l=1 n A kl 2 Let A be normal (A A = AA ) and upper triangular (A ij = 0 if i > j) Then l=1 A 11 2 = n A l1 2 = (A A) 11 = (AA ) 11 = n A 1l 2 = A n A 1l 2, l=1 l=1 l=2 which shows that A 1l = 0 for l 2 From this result ( ), next A 22 2 ( ) = n A l2 2 = (A A) 22 = (AA ) 22 = l=1 n A 2l 2 = A l=1 n A 2l 2 l=3 shows that A 2l = 0 for l 3 Inductively, this yields A ij = 0 whenever j > i (and we already had A ji = 0 here) Hence A is a diagonal matrix! Result: So, normal triangular matrices are actually diagonal! (A trivial result: symmetric triangular matrices are diagonal and real! Think about that!) Remark This previous result soon becomes a powerful tool Namely, we will show that all square matrices can be unitarily triangulated; hence all normal matrices can be unitarily diagonalized! (Especially, all symmetric matrices can be unitarily diagonalized) 54

55 Gram Schmidt process reformulated Let matrix S = [S 1 S n ] C n n be invertible, with columns S k C n 1 The Gram Schmidt process gives unitary U = [U 1 U n ] C n n as follows: Let U 1 := S 1 / S 1, (42) and for k 2 then where U k := V k V k, (43) k 1 V k := S k S k, U j U j (44) j=1 It is trivial that U k, U k = 1 Also, U k, U l = 0 for k l, because if k > l then k V k, U l = S k, U l S k, U j U j, U l = S k, U l S k, U l = 0 j=1 Notice that for all k {1,, n} here span{s 1,, S k } = span{u 1,, U k } This Gram Schmidt process is numerically unstable (round-off errors accumulate), but there are ways to stabilize the process Example Let S = and [ ] 4 6 Then 3 2 U 1 = S 1 / S 1 = [ ] 4 / = [ ] 4/5, 3/5 V 2 = S 2 S 1, U 1 U [ ] [ ] [ 1 ] [ ] 6 6 4/5 4/5 =, 2 2 3/5 3/5 [ ] so U 2 = V 2 3/5 = V 2 We have 4/5 U = [U 1 U 2 ] = = [ ] [ ] 6 4/ /5 [ ] 4/5 3/5 3/5 4/5 = [ ] 6/5, 8/5 55

56 S 1 S 2, U 1 U 1 S2 U 1 U 2 V 2 Figure 35: Gram Schmidt: from invertible S = [S 1 S n ] C n n to unitary U = [U 1 U n ] C n n such that span{s 1,, S k } = span{u 1,, U k } Unitary triangulation of square matrices For a matrix A n C n n, we next find a unitary matrix U n C n n and an upper triangular matrix Λ n C n n such that A n = U n Λ n U n (45) This presentation (45) is trivial when n = 1, so let us assume n > 1 We prove (45) by induction, by reducing case n to case n 1 Take an eigenvalue λ C with a normalized eigenvector v C n 1 : A n v = λv, v = 1 By Gram Schmidt, find unitary V C n n with first column v So [ ] λ w A n = V V O A n 1 for some w C 1 (n 1), where O R (n 1) 1 is the zero column vector, and A n 1 = U n 1 Λ n 1 U n 1 C (n 1) (n 1) by case n 1 of (45) Let [ ] 1 O U n := V, O U n 1 [ ] λ wun 1 Λ n = O Λ n 1 Then A n = U n Λ n U n, where U n is unitary and Λ n upper triangular 56

57 A(U 1 ) = λ 1 U 1 U 2 U 1 A:C n C n A(U 2 ) = λ 2 U 2 O C n O = AO C n U = [U 1 U n ] U = [U 1 U n ] I 2 = U (U 2 ) C n Λ:C n C n λ 2 I 2 C n O C n I 1 = U (U 1 ) C n O = ΛO C n λ 1 I 1 C n Figure 36: Idea of unitary diagonalization Λ = U AU of normal A C n n Equivalently, this means A = UΛU Unitary operations U and U preserve distances and angles Unitary diagonalization of normal matrices Thus, for any A C n n there is unitary triangulation A = UΛU, where U C n n is unitary and Λ C n n is upper triangular It is easy to see that here Λ is normal if and only if A is normal Because normal triangular matrices are diagonal (see page 54), we get: Normal matrices can be diagonalized by unitary matrices! More precisely: Theorem Conditions (1) and (2) are equivalent: (1) A C n n is normal (that is, A A = AA ) (2) A = UΛU for unitary U C n n and diagonal Λ C n n Remark: In this result on the unitary diagonalization A = UΛU, it is easy to see that A = A if and only if Λ = Λ (Why?) So, a normal matrix is symmetric if and only if its eigenvalues are real However, there are nonnormal diagonalizable matrices with real eigenvalues: see the example on page 47 57

Math Linear Algebra Final Exam Review Sheet

Math Linear Algebra Final Exam Review Sheet Math 15-1 Linear Algebra Final Exam Review Sheet Vector Operations Vector addition is a component-wise operation. Two vectors v and w may be added together as long as they contain the same number n of

More information

CS 246 Review of Linear Algebra 01/17/19

CS 246 Review of Linear Algebra 01/17/19 1 Linear algebra In this section we will discuss vectors and matrices. We denote the (i, j)th entry of a matrix A as A ij, and the ith entry of a vector as v i. 1.1 Vectors and vector operations A vector

More information

1. General Vector Spaces

1. General Vector Spaces 1.1. Vector space axioms. 1. General Vector Spaces Definition 1.1. Let V be a nonempty set of objects on which the operations of addition and scalar multiplication are defined. By addition we mean a rule

More information

Math Linear Algebra II. 1. Inner Products and Norms

Math Linear Algebra II. 1. Inner Products and Norms Math 342 - Linear Algebra II Notes 1. Inner Products and Norms One knows from a basic introduction to vectors in R n Math 254 at OSU) that the length of a vector x = x 1 x 2... x n ) T R n, denoted x,

More information

Review problems for MA 54, Fall 2004.

Review problems for MA 54, Fall 2004. Review problems for MA 54, Fall 2004. Below are the review problems for the final. They are mostly homework problems, or very similar. If you are comfortable doing these problems, you should be fine on

More information

Ir O D = D = ( ) Section 2.6 Example 1. (Bottom of page 119) dim(v ) = dim(l(v, W )) = dim(v ) dim(f ) = dim(v )

Ir O D = D = ( ) Section 2.6 Example 1. (Bottom of page 119) dim(v ) = dim(l(v, W )) = dim(v ) dim(f ) = dim(v ) Section 3.2 Theorem 3.6. Let A be an m n matrix of rank r. Then r m, r n, and, by means of a finite number of elementary row and column operations, A can be transformed into the matrix ( ) Ir O D = 1 O

More information

Linear Algebra Primer

Linear Algebra Primer Linear Algebra Primer David Doria daviddoria@gmail.com Wednesday 3 rd December, 2008 Contents Why is it called Linear Algebra? 4 2 What is a Matrix? 4 2. Input and Output.....................................

More information

HOMEWORK PROBLEMS FROM STRANG S LINEAR ALGEBRA AND ITS APPLICATIONS (4TH EDITION)

HOMEWORK PROBLEMS FROM STRANG S LINEAR ALGEBRA AND ITS APPLICATIONS (4TH EDITION) HOMEWORK PROBLEMS FROM STRANG S LINEAR ALGEBRA AND ITS APPLICATIONS (4TH EDITION) PROFESSOR STEVEN MILLER: BROWN UNIVERSITY: SPRING 2007 1. CHAPTER 1: MATRICES AND GAUSSIAN ELIMINATION Page 9, # 3: Describe

More information

DS-GA 1002 Lecture notes 0 Fall Linear Algebra. These notes provide a review of basic concepts in linear algebra.

DS-GA 1002 Lecture notes 0 Fall Linear Algebra. These notes provide a review of basic concepts in linear algebra. DS-GA 1002 Lecture notes 0 Fall 2016 Linear Algebra These notes provide a review of basic concepts in linear algebra. 1 Vector spaces You are no doubt familiar with vectors in R 2 or R 3, i.e. [ ] 1.1

More information

IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET

IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET This is a (not quite comprehensive) list of definitions and theorems given in Math 1553. Pay particular attention to the ones in red. Study Tip For each

More information

Math 102, Winter Final Exam Review. Chapter 1. Matrices and Gaussian Elimination

Math 102, Winter Final Exam Review. Chapter 1. Matrices and Gaussian Elimination Math 0, Winter 07 Final Exam Review Chapter. Matrices and Gaussian Elimination { x + x =,. Different forms of a system of linear equations. Example: The x + 4x = 4. [ ] [ ] [ ] vector form (or the column

More information

IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET

IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET IMPORTANT DEFINITIONS AND THEOREMS REFERENCE SHEET This is a (not quite comprehensive) list of definitions and theorems given in Math 1553. Pay particular attention to the ones in red. Study Tip For each

More information

Elementary linear algebra

Elementary linear algebra Chapter 1 Elementary linear algebra 1.1 Vector spaces Vector spaces owe their importance to the fact that so many models arising in the solutions of specific problems turn out to be vector spaces. The

More information

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces.

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces. Math 350 Fall 2011 Notes about inner product spaces In this notes we state and prove some important properties of inner product spaces. First, recall the dot product on R n : if x, y R n, say x = (x 1,...,

More information

3 Matrix Algebra. 3.1 Operations on matrices

3 Matrix Algebra. 3.1 Operations on matrices 3 Matrix Algebra A matrix is a rectangular array of numbers; it is of size m n if it has m rows and n columns. A 1 n matrix is a row vector; an m 1 matrix is a column vector. For example: 1 5 3 5 3 5 8

More information

Linear Algebra March 16, 2019

Linear Algebra March 16, 2019 Linear Algebra March 16, 2019 2 Contents 0.1 Notation................................ 4 1 Systems of linear equations, and matrices 5 1.1 Systems of linear equations..................... 5 1.2 Augmented

More information

Linear Algebra M1 - FIB. Contents: 5. Matrices, systems of linear equations and determinants 6. Vector space 7. Linear maps 8.

Linear Algebra M1 - FIB. Contents: 5. Matrices, systems of linear equations and determinants 6. Vector space 7. Linear maps 8. Linear Algebra M1 - FIB Contents: 5 Matrices, systems of linear equations and determinants 6 Vector space 7 Linear maps 8 Diagonalization Anna de Mier Montserrat Maureso Dept Matemàtica Aplicada II Translation:

More information

MATH 240 Spring, Chapter 1: Linear Equations and Matrices

MATH 240 Spring, Chapter 1: Linear Equations and Matrices MATH 240 Spring, 2006 Chapter Summaries for Kolman / Hill, Elementary Linear Algebra, 8th Ed. Sections 1.1 1.6, 2.1 2.2, 3.2 3.8, 4.3 4.5, 5.1 5.3, 5.5, 6.1 6.5, 7.1 7.2, 7.4 DEFINITIONS Chapter 1: Linear

More information

2. Every linear system with the same number of equations as unknowns has a unique solution.

2. Every linear system with the same number of equations as unknowns has a unique solution. 1. For matrices A, B, C, A + B = A + C if and only if A = B. 2. Every linear system with the same number of equations as unknowns has a unique solution. 3. Every linear system with the same number of equations

More information

Math113: Linear Algebra. Beifang Chen

Math113: Linear Algebra. Beifang Chen Math3: Linear Algebra Beifang Chen Spring 26 Contents Systems of Linear Equations 3 Systems of Linear Equations 3 Linear Systems 3 2 Geometric Interpretation 3 3 Matrices of Linear Systems 4 4 Elementary

More information

Knowledge Discovery and Data Mining 1 (VO) ( )

Knowledge Discovery and Data Mining 1 (VO) ( ) Knowledge Discovery and Data Mining 1 (VO) (707.003) Review of Linear Algebra Denis Helic KTI, TU Graz Oct 9, 2014 Denis Helic (KTI, TU Graz) KDDM1 Oct 9, 2014 1 / 74 Big picture: KDDM Probability Theory

More information

Conceptual Questions for Review

Conceptual Questions for Review Conceptual Questions for Review Chapter 1 1.1 Which vectors are linear combinations of v = (3, 1) and w = (4, 3)? 1.2 Compare the dot product of v = (3, 1) and w = (4, 3) to the product of their lengths.

More information

Introduction to Linear Algebra, Second Edition, Serge Lange

Introduction to Linear Algebra, Second Edition, Serge Lange Introduction to Linear Algebra, Second Edition, Serge Lange Chapter I: Vectors R n defined. Addition and scalar multiplication in R n. Two geometric interpretations for a vector: point and displacement.

More information

Basic Elements of Linear Algebra

Basic Elements of Linear Algebra A Basic Review of Linear Algebra Nick West nickwest@stanfordedu September 16, 2010 Part I Basic Elements of Linear Algebra Although the subject of linear algebra is much broader than just vectors and matrices,

More information

A Brief Outline of Math 355

A Brief Outline of Math 355 A Brief Outline of Math 355 Lecture 1 The geometry of linear equations; elimination with matrices A system of m linear equations with n unknowns can be thought of geometrically as m hyperplanes intersecting

More information

Linear Algebra. Matrices Operations. Consider, for example, a system of equations such as x + 2y z + 4w = 0, 3x 4y + 2z 6w = 0, x 3y 2z + w = 0.

Linear Algebra. Matrices Operations. Consider, for example, a system of equations such as x + 2y z + 4w = 0, 3x 4y + 2z 6w = 0, x 3y 2z + w = 0. Matrices Operations Linear Algebra Consider, for example, a system of equations such as x + 2y z + 4w = 0, 3x 4y + 2z 6w = 0, x 3y 2z + w = 0 The rectangular array 1 2 1 4 3 4 2 6 1 3 2 1 in which the

More information

Introduction to Matrix Algebra

Introduction to Matrix Algebra Introduction to Matrix Algebra August 18, 2010 1 Vectors 1.1 Notations A p-dimensional vector is p numbers put together. Written as x 1 x =. x p. When p = 1, this represents a point in the line. When p

More information

Linear Algebra Highlights

Linear Algebra Highlights Linear Algebra Highlights Chapter 1 A linear equation in n variables is of the form a 1 x 1 + a 2 x 2 + + a n x n. We can have m equations in n variables, a system of linear equations, which we want to

More information

1 Matrices and Systems of Linear Equations. a 1n a 2n

1 Matrices and Systems of Linear Equations. a 1n a 2n March 31, 2013 16-1 16. Systems of Linear Equations 1 Matrices and Systems of Linear Equations An m n matrix is an array A = (a ij ) of the form a 11 a 21 a m1 a 1n a 2n... a mn where each a ij is a real

More information

Linear Algebra- Final Exam Review

Linear Algebra- Final Exam Review Linear Algebra- Final Exam Review. Let A be invertible. Show that, if v, v, v 3 are linearly independent vectors, so are Av, Av, Av 3. NOTE: It should be clear from your answer that you know the definition.

More information

Solution to Homework 1

Solution to Homework 1 Solution to Homework Sec 2 (a) Yes It is condition (VS 3) (b) No If x, y are both zero vectors Then by condition (VS 3) x = x + y = y (c) No Let e be the zero vector We have e = 2e (d) No It will be false

More information

Math 3108: Linear Algebra

Math 3108: Linear Algebra Math 3108: Linear Algebra Instructor: Jason Murphy Department of Mathematics and Statistics Missouri University of Science and Technology 1 / 323 Contents. Chapter 1. Slides 3 70 Chapter 2. Slides 71 118

More information

Final Review Sheet. B = (1, 1 + 3x, 1 + x 2 ) then 2 + 3x + 6x 2

Final Review Sheet. B = (1, 1 + 3x, 1 + x 2 ) then 2 + 3x + 6x 2 Final Review Sheet The final will cover Sections Chapters 1,2,3 and 4, as well as sections 5.1-5.4, 6.1-6.2 and 7.1-7.3 from chapters 5,6 and 7. This is essentially all material covered this term. Watch

More information

1 Last time: least-squares problems

1 Last time: least-squares problems MATH Linear algebra (Fall 07) Lecture Last time: least-squares problems Definition. If A is an m n matrix and b R m, then a least-squares solution to the linear system Ax = b is a vector x R n such that

More information

LINEAR ALGEBRA REVIEW

LINEAR ALGEBRA REVIEW LINEAR ALGEBRA REVIEW JC Stuff you should know for the exam. 1. Basics on vector spaces (1) F n is the set of all n-tuples (a 1,... a n ) with a i F. It forms a VS with the operations of + and scalar multiplication

More information

OHSx XM511 Linear Algebra: Solutions to Online True/False Exercises

OHSx XM511 Linear Algebra: Solutions to Online True/False Exercises This document gives the solutions to all of the online exercises for OHSx XM511. The section ( ) numbers refer to the textbook. TYPE I are True/False. Answers are in square brackets [. Lecture 02 ( 1.1)

More information

The following definition is fundamental.

The following definition is fundamental. 1. Some Basics from Linear Algebra With these notes, I will try and clarify certain topics that I only quickly mention in class. First and foremost, I will assume that you are familiar with many basic

More information

[Disclaimer: This is not a complete list of everything you need to know, just some of the topics that gave people difficulty.]

[Disclaimer: This is not a complete list of everything you need to know, just some of the topics that gave people difficulty.] Math 43 Review Notes [Disclaimer: This is not a complete list of everything you need to know, just some of the topics that gave people difficulty Dot Product If v (v, v, v 3 and w (w, w, w 3, then the

More information

Math Camp Lecture 4: Linear Algebra. Xiao Yu Wang. Aug 2010 MIT. Xiao Yu Wang (MIT) Math Camp /10 1 / 88

Math Camp Lecture 4: Linear Algebra. Xiao Yu Wang. Aug 2010 MIT. Xiao Yu Wang (MIT) Math Camp /10 1 / 88 Math Camp 2010 Lecture 4: Linear Algebra Xiao Yu Wang MIT Aug 2010 Xiao Yu Wang (MIT) Math Camp 2010 08/10 1 / 88 Linear Algebra Game Plan Vector Spaces Linear Transformations and Matrices Determinant

More information

Equality: Two matrices A and B are equal, i.e., A = B if A and B have the same order and the entries of A and B are the same.

Equality: Two matrices A and B are equal, i.e., A = B if A and B have the same order and the entries of A and B are the same. Introduction Matrix Operations Matrix: An m n matrix A is an m-by-n array of scalars from a field (for example real numbers) of the form a a a n a a a n A a m a m a mn The order (or size) of A is m n (read

More information

Problem Set (T) If A is an m n matrix, B is an n p matrix and D is a p s matrix, then show

Problem Set (T) If A is an m n matrix, B is an n p matrix and D is a p s matrix, then show MTH 0: Linear Algebra Department of Mathematics and Statistics Indian Institute of Technology - Kanpur Problem Set Problems marked (T) are for discussions in Tutorial sessions (T) If A is an m n matrix,

More information

October 25, 2013 INNER PRODUCT SPACES

October 25, 2013 INNER PRODUCT SPACES October 25, 2013 INNER PRODUCT SPACES RODICA D. COSTIN Contents 1. Inner product 2 1.1. Inner product 2 1.2. Inner product spaces 4 2. Orthogonal bases 5 2.1. Existence of an orthogonal basis 7 2.2. Orthogonal

More information

Linear Algebra Lecture Notes-II

Linear Algebra Lecture Notes-II Linear Algebra Lecture Notes-II Vikas Bist Department of Mathematics Panjab University, Chandigarh-64 email: bistvikas@gmail.com Last revised on March 5, 8 This text is based on the lectures delivered

More information

MATH 106 LINEAR ALGEBRA LECTURE NOTES

MATH 106 LINEAR ALGEBRA LECTURE NOTES MATH 6 LINEAR ALGEBRA LECTURE NOTES FALL - These Lecture Notes are not in a final form being still subject of improvement Contents Systems of linear equations and matrices 5 Introduction to systems of

More information

Linear Algebra. Workbook

Linear Algebra. Workbook Linear Algebra Workbook Paul Yiu Department of Mathematics Florida Atlantic University Last Update: November 21 Student: Fall 2011 Checklist Name: A B C D E F F G H I J 1 2 3 4 5 6 7 8 9 10 xxx xxx xxx

More information

Matrices and Linear Algebra

Matrices and Linear Algebra Contents Quantitative methods for Economics and Business University of Ferrara Academic year 2017-2018 Contents 1 Basics 2 3 4 5 Contents 1 Basics 2 3 4 5 Contents 1 Basics 2 3 4 5 Contents 1 Basics 2

More information

Linear Algebra Review

Linear Algebra Review Chapter 1 Linear Algebra Review It is assumed that you have had a course in linear algebra, and are familiar with matrix multiplication, eigenvectors, etc. I will review some of these terms here, but quite

More information

1 Matrices and Systems of Linear Equations

1 Matrices and Systems of Linear Equations March 3, 203 6-6. Systems of Linear Equations Matrices and Systems of Linear Equations An m n matrix is an array A = a ij of the form a a n a 2 a 2n... a m a mn where each a ij is a real or complex number.

More information

LINEAR ALGEBRA SUMMARY SHEET.

LINEAR ALGEBRA SUMMARY SHEET. LINEAR ALGEBRA SUMMARY SHEET RADON ROSBOROUGH https://intuitiveexplanationscom/linear-algebra-summary-sheet/ This document is a concise collection of many of the important theorems of linear algebra, organized

More information

CS 143 Linear Algebra Review

CS 143 Linear Algebra Review CS 143 Linear Algebra Review Stefan Roth September 29, 2003 Introductory Remarks This review does not aim at mathematical rigor very much, but instead at ease of understanding and conciseness. Please see

More information

4. Determinants.

4. Determinants. 4. Determinants 4.1. Determinants; Cofactor Expansion Determinants of 2 2 and 3 3 Matrices 2 2 determinant 4.1. Determinants; Cofactor Expansion Determinants of 2 2 and 3 3 Matrices 3 3 determinant 4.1.

More information

LINEAR ALGEBRA MICHAEL PENKAVA

LINEAR ALGEBRA MICHAEL PENKAVA LINEAR ALGEBRA MICHAEL PENKAVA 1. Linear Maps Definition 1.1. If V and W are vector spaces over the same field K, then a map λ : V W is called a linear map if it satisfies the two conditions below: (1)

More information

Math 18, Linear Algebra, Lecture C00, Spring 2017 Review and Practice Problems for Final Exam

Math 18, Linear Algebra, Lecture C00, Spring 2017 Review and Practice Problems for Final Exam Math 8, Linear Algebra, Lecture C, Spring 7 Review and Practice Problems for Final Exam. The augmentedmatrix of a linear system has been transformed by row operations into 5 4 8. Determine if the system

More information

ELEMENTARY LINEAR ALGEBRA WITH APPLICATIONS. 1. Linear Equations and Matrices

ELEMENTARY LINEAR ALGEBRA WITH APPLICATIONS. 1. Linear Equations and Matrices ELEMENTARY LINEAR ALGEBRA WITH APPLICATIONS KOLMAN & HILL NOTES BY OTTO MUTZBAUER 11 Systems of Linear Equations 1 Linear Equations and Matrices Numbers in our context are either real numbers or complex

More information

MATRICES ARE SIMILAR TO TRIANGULAR MATRICES

MATRICES ARE SIMILAR TO TRIANGULAR MATRICES MATRICES ARE SIMILAR TO TRIANGULAR MATRICES 1 Complex matrices Recall that the complex numbers are given by a + ib where a and b are real and i is the imaginary unity, ie, i 2 = 1 In what we describe below,

More information

Extra Problems for Math 2050 Linear Algebra I

Extra Problems for Math 2050 Linear Algebra I Extra Problems for Math 5 Linear Algebra I Find the vector AB and illustrate with a picture if A = (,) and B = (,4) Find B, given A = (,4) and [ AB = A = (,4) and [ AB = 8 If possible, express x = 7 as

More information

Linear Algebra in Actuarial Science: Slides to the lecture

Linear Algebra in Actuarial Science: Slides to the lecture Linear Algebra in Actuarial Science: Slides to the lecture Fall Semester 2010/2011 Linear Algebra is a Tool-Box Linear Equation Systems Discretization of differential equations: solving linear equations

More information

Matrix Theory. A.Holst, V.Ufnarovski

Matrix Theory. A.Holst, V.Ufnarovski Matrix Theory AHolst, VUfnarovski 55 HINTS AND ANSWERS 9 55 Hints and answers There are two different approaches In the first one write A as a block of rows and note that in B = E ij A all rows different

More information

MTH 2032 SemesterII

MTH 2032 SemesterII MTH 202 SemesterII 2010-11 Linear Algebra Worked Examples Dr. Tony Yee Department of Mathematics and Information Technology The Hong Kong Institute of Education December 28, 2011 ii Contents Table of Contents

More information

NORMS ON SPACE OF MATRICES

NORMS ON SPACE OF MATRICES NORMS ON SPACE OF MATRICES. Operator Norms on Space of linear maps Let A be an n n real matrix and x 0 be a vector in R n. We would like to use the Picard iteration method to solve for the following system

More information

LINEAR ALGEBRA W W L CHEN

LINEAR ALGEBRA W W L CHEN LINEAR ALGEBRA W W L CHEN c W W L Chen, 1997, 2008. This chapter is available free to all individuals, on the understanding that it is not to be used for financial gain, and may be downloaded and/or photocopied,

More information

Quantum Computing Lecture 2. Review of Linear Algebra

Quantum Computing Lecture 2. Review of Linear Algebra Quantum Computing Lecture 2 Review of Linear Algebra Maris Ozols Linear algebra States of a quantum system form a vector space and their transformations are described by linear operators Vector spaces

More information

NOTES FOR LINEAR ALGEBRA 133

NOTES FOR LINEAR ALGEBRA 133 NOTES FOR LINEAR ALGEBRA 33 William J Anderson McGill University These are not official notes for Math 33 identical to the notes projected in class They are intended for Anderson s section 4, and are 2

More information

6 Inner Product Spaces

6 Inner Product Spaces Lectures 16,17,18 6 Inner Product Spaces 6.1 Basic Definition Parallelogram law, the ability to measure angle between two vectors and in particular, the concept of perpendicularity make the euclidean space

More information

Vectors and matrices: matrices (Version 2) This is a very brief summary of my lecture notes.

Vectors and matrices: matrices (Version 2) This is a very brief summary of my lecture notes. Vectors and matrices: matrices (Version 2) This is a very brief summary of my lecture notes Matrices and linear equations A matrix is an m-by-n array of numbers A = a 11 a 12 a 13 a 1n a 21 a 22 a 23 a

More information

Algebra II. Paulius Drungilas and Jonas Jankauskas

Algebra II. Paulius Drungilas and Jonas Jankauskas Algebra II Paulius Drungilas and Jonas Jankauskas Contents 1. Quadratic forms 3 What is quadratic form? 3 Change of variables. 3 Equivalence of quadratic forms. 4 Canonical form. 4 Normal form. 7 Positive

More information

Unit 3: Matrices. Juan Luis Melero and Eduardo Eyras. September 2018

Unit 3: Matrices. Juan Luis Melero and Eduardo Eyras. September 2018 Unit 3: Matrices Juan Luis Melero and Eduardo Eyras September 2018 1 Contents 1 Matrices and operations 4 1.1 Definition of a matrix....................... 4 1.2 Addition and subtraction of matrices..............

More information

Math Bootcamp An p-dimensional vector is p numbers put together. Written as. x 1 x =. x p

Math Bootcamp An p-dimensional vector is p numbers put together. Written as. x 1 x =. x p Math Bootcamp 2012 1 Review of matrix algebra 1.1 Vectors and rules of operations An p-dimensional vector is p numbers put together. Written as x 1 x =. x p. When p = 1, this represents a point in the

More information

LINEAR ALGEBRA WITH APPLICATIONS

LINEAR ALGEBRA WITH APPLICATIONS SEVENTH EDITION LINEAR ALGEBRA WITH APPLICATIONS Instructor s Solutions Manual Steven J. Leon PREFACE This solutions manual is designed to accompany the seventh edition of Linear Algebra with Applications

More information

Recall: Dot product on R 2 : u v = (u 1, u 2 ) (v 1, v 2 ) = u 1 v 1 + u 2 v 2, u u = u u 2 2 = u 2. Geometric Meaning:

Recall: Dot product on R 2 : u v = (u 1, u 2 ) (v 1, v 2 ) = u 1 v 1 + u 2 v 2, u u = u u 2 2 = u 2. Geometric Meaning: Recall: Dot product on R 2 : u v = (u 1, u 2 ) (v 1, v 2 ) = u 1 v 1 + u 2 v 2, u u = u 2 1 + u 2 2 = u 2. Geometric Meaning: u v = u v cos θ. u θ v 1 Reason: The opposite side is given by u v. u v 2 =

More information

Chapter 2 Notes, Linear Algebra 5e Lay

Chapter 2 Notes, Linear Algebra 5e Lay Contents.1 Operations with Matrices..................................1.1 Addition and Subtraction.............................1. Multiplication by a scalar............................ 3.1.3 Multiplication

More information

x 3y 2z = 6 1.2) 2x 4y 3z = 8 3x + 6y + 8z = 5 x + 3y 2z + 5t = 4 1.5) 2x + 8y z + 9t = 9 3x + 5y 12z + 17t = 7

x 3y 2z = 6 1.2) 2x 4y 3z = 8 3x + 6y + 8z = 5 x + 3y 2z + 5t = 4 1.5) 2x + 8y z + 9t = 9 3x + 5y 12z + 17t = 7 Linear Algebra and its Applications-Lab 1 1) Use Gaussian elimination to solve the following systems x 1 + x 2 2x 3 + 4x 4 = 5 1.1) 2x 1 + 2x 2 3x 3 + x 4 = 3 3x 1 + 3x 2 4x 3 2x 4 = 1 x + y + 2z = 4 1.4)

More information

Matrix Algebra Determinant, Inverse matrix. Matrices. A. Fabretti. Mathematics 2 A.Y. 2015/2016. A. Fabretti Matrices

Matrix Algebra Determinant, Inverse matrix. Matrices. A. Fabretti. Mathematics 2 A.Y. 2015/2016. A. Fabretti Matrices Matrices A. Fabretti Mathematics 2 A.Y. 2015/2016 Table of contents Matrix Algebra Determinant Inverse Matrix Introduction A matrix is a rectangular array of numbers. The size of a matrix is indicated

More information

Math 4A Notes. Written by Victoria Kala Last updated June 11, 2017

Math 4A Notes. Written by Victoria Kala Last updated June 11, 2017 Math 4A Notes Written by Victoria Kala vtkala@math.ucsb.edu Last updated June 11, 2017 Systems of Linear Equations A linear equation is an equation that can be written in the form a 1 x 1 + a 2 x 2 +...

More information

Stat 159/259: Linear Algebra Notes

Stat 159/259: Linear Algebra Notes Stat 159/259: Linear Algebra Notes Jarrod Millman November 16, 2015 Abstract These notes assume you ve taken a semester of undergraduate linear algebra. In particular, I assume you are familiar with the

More information

Review of linear algebra

Review of linear algebra Review of linear algebra 1 Vectors and matrices We will just touch very briefly on certain aspects of linear algebra, most of which should be familiar. Recall that we deal with vectors, i.e. elements of

More information

LINEAR ALGEBRA BOOT CAMP WEEK 4: THE SPECTRAL THEOREM

LINEAR ALGEBRA BOOT CAMP WEEK 4: THE SPECTRAL THEOREM LINEAR ALGEBRA BOOT CAMP WEEK 4: THE SPECTRAL THEOREM Unless otherwise stated, all vector spaces in this worksheet are finite dimensional and the scalar field F is R or C. Definition 1. A linear operator

More information

GQE ALGEBRA PROBLEMS

GQE ALGEBRA PROBLEMS GQE ALGEBRA PROBLEMS JAKOB STREIPEL Contents. Eigenthings 2. Norms, Inner Products, Orthogonality, and Such 6 3. Determinants, Inverses, and Linear (In)dependence 4. (Invariant) Subspaces 3 Throughout

More information

1 9/5 Matrices, vectors, and their applications

1 9/5 Matrices, vectors, and their applications 1 9/5 Matrices, vectors, and their applications Algebra: study of objects and operations on them. Linear algebra: object: matrices and vectors. operations: addition, multiplication etc. Algorithms/Geometric

More information

Chapter 6: Orthogonality

Chapter 6: Orthogonality Chapter 6: Orthogonality (Last Updated: November 7, 7) These notes are derived primarily from Linear Algebra and its applications by David Lay (4ed). A few theorems have been moved around.. Inner products

More information

Linear algebra 2. Yoav Zemel. March 1, 2012

Linear algebra 2. Yoav Zemel. March 1, 2012 Linear algebra 2 Yoav Zemel March 1, 2012 These notes were written by Yoav Zemel. The lecturer, Shmuel Berger, should not be held responsible for any mistake. Any comments are welcome at zamsh7@gmail.com.

More information

Linear Algebra II. Ulrike Tillmann. January 4, 2018

Linear Algebra II. Ulrike Tillmann. January 4, 2018 Linear Algebra II Ulrike Tillmann January 4, 208 This course is a continuation of Linear Algebra I and will foreshadow much of what will be discussed in more detail in the Linear Algebra course in Part

More information

Elementary Linear Algebra

Elementary Linear Algebra Matrices J MUSCAT Elementary Linear Algebra Matrices Definition Dr J Muscat 2002 A matrix is a rectangular array of numbers, arranged in rows and columns a a 2 a 3 a n a 2 a 22 a 23 a 2n A = a m a mn We

More information

Math 108b: Notes on the Spectral Theorem

Math 108b: Notes on the Spectral Theorem Math 108b: Notes on the Spectral Theorem From section 6.3, we know that every linear operator T on a finite dimensional inner product space V has an adjoint. (T is defined as the unique linear operator

More information

Linear Algebra: Lecture Notes. Dr Rachel Quinlan School of Mathematics, Statistics and Applied Mathematics NUI Galway

Linear Algebra: Lecture Notes. Dr Rachel Quinlan School of Mathematics, Statistics and Applied Mathematics NUI Galway Linear Algebra: Lecture Notes Dr Rachel Quinlan School of Mathematics, Statistics and Applied Mathematics NUI Galway November 6, 23 Contents Systems of Linear Equations 2 Introduction 2 2 Elementary Row

More information

EXERCISE SET 5.1. = (kx + kx + k, ky + ky + k ) = (kx + kx + 1, ky + ky + 1) = ((k + )x + 1, (k + )y + 1)

EXERCISE SET 5.1. = (kx + kx + k, ky + ky + k ) = (kx + kx + 1, ky + ky + 1) = ((k + )x + 1, (k + )y + 1) EXERCISE SET 5. 6. The pair (, 2) is in the set but the pair ( )(, 2) = (, 2) is not because the first component is negative; hence Axiom 6 fails. Axiom 5 also fails. 8. Axioms, 2, 3, 6, 9, and are easily

More information

Notes on Linear Algebra

Notes on Linear Algebra 1 Notes on Linear Algebra Jean Walrand August 2005 I INTRODUCTION Linear Algebra is the theory of linear transformations Applications abound in estimation control and Markov chains You should be familiar

More information

Linear Algebra Massoud Malek

Linear Algebra Massoud Malek CSUEB Linear Algebra Massoud Malek Inner Product and Normed Space In all that follows, the n n identity matrix is denoted by I n, the n n zero matrix by Z n, and the zero vector by θ n An inner product

More information

Lecture 2: Linear operators

Lecture 2: Linear operators Lecture 2: Linear operators Rajat Mittal IIT Kanpur The mathematical formulation of Quantum computing requires vector spaces and linear operators So, we need to be comfortable with linear algebra to study

More information

Calculating determinants for larger matrices

Calculating determinants for larger matrices Day 26 Calculating determinants for larger matrices We now proceed to define det A for n n matrices A As before, we are looking for a function of A that satisfies the product formula det(ab) = det A det

More information

j=1 x j p, if 1 p <, x i ξ : x i < ξ} 0 as p.

j=1 x j p, if 1 p <, x i ξ : x i < ξ} 0 as p. LINEAR ALGEBRA Fall 203 The final exam Almost all of the problems solved Exercise Let (V, ) be a normed vector space. Prove x y x y for all x, y V. Everybody knows how to do this! Exercise 2 If V is a

More information

Math 520 Exam 2 Topic Outline Sections 1 3 (Xiao/Dumas/Liaw) Spring 2008

Math 520 Exam 2 Topic Outline Sections 1 3 (Xiao/Dumas/Liaw) Spring 2008 Math 520 Exam 2 Topic Outline Sections 1 3 (Xiao/Dumas/Liaw) Spring 2008 Exam 2 will be held on Tuesday, April 8, 7-8pm in 117 MacMillan What will be covered The exam will cover material from the lectures

More information

Math 291-2: Lecture Notes Northwestern University, Winter 2016

Math 291-2: Lecture Notes Northwestern University, Winter 2016 Math 291-2: Lecture Notes Northwestern University, Winter 2016 Written by Santiago Cañez These are lecture notes for Math 291-2, the second quarter of MENU: Intensive Linear Algebra and Multivariable Calculus,

More information

Chapter 6 Inner product spaces

Chapter 6 Inner product spaces Chapter 6 Inner product spaces 6.1 Inner products and norms Definition 1 Let V be a vector space over F. An inner product on V is a function, : V V F such that the following conditions hold. x+z,y = x,y

More information

G1110 & 852G1 Numerical Linear Algebra

G1110 & 852G1 Numerical Linear Algebra The University of Sussex Department of Mathematics G & 85G Numerical Linear Algebra Lecture Notes Autumn Term Kerstin Hesse (w aw S w a w w (w aw H(wa = (w aw + w Figure : Geometric explanation of the

More information

Dot Products. K. Behrend. April 3, Abstract A short review of some basic facts on the dot product. Projections. The spectral theorem.

Dot Products. K. Behrend. April 3, Abstract A short review of some basic facts on the dot product. Projections. The spectral theorem. Dot Products K. Behrend April 3, 008 Abstract A short review of some basic facts on the dot product. Projections. The spectral theorem. Contents The dot product 3. Length of a vector........................

More information

MATH Topics in Applied Mathematics Lecture 12: Evaluation of determinants. Cross product.

MATH Topics in Applied Mathematics Lecture 12: Evaluation of determinants. Cross product. MATH 311-504 Topics in Applied Mathematics Lecture 12: Evaluation of determinants. Cross product. Determinant is a scalar assigned to each square matrix. Notation. The determinant of a matrix A = (a ij

More information

Discrete Math, Second Problem Set (June 24)

Discrete Math, Second Problem Set (June 24) Discrete Math, Second Problem Set (June 24) REU 2003 Instructor: Laszlo Babai Scribe: D Jeremy Copeland 1 Number Theory Remark 11 For an arithmetic progression, a 0, a 1 = a 0 +d, a 2 = a 0 +2d, to have

More information

MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators.

MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators. MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators. Adjoint operator and adjoint matrix Given a linear operator L on an inner product space V, the adjoint of L is a transformation

More information

Eigenvalues and Eigenvectors

Eigenvalues and Eigenvectors /88 Chia-Ping Chen Department of Computer Science and Engineering National Sun Yat-sen University Linear Algebra Eigenvalue Problem /88 Eigenvalue Equation By definition, the eigenvalue equation for matrix

More information