Linear Algebra Chapter 1. Vectors in R n and C n , spatial vectors A vector v is a member of a Vector Space R n ,C n . If u ,v V where V is a vector space, and , are scalars (numbers) then ( u v ) u v , ( )u u u , ( u ) ( )u are all in V . A vector is an ordered collection of numbers (quantities), an array. e.g. v [ 2.3 - 3.5 0.1] v ( on on off on on off) v ( blue green blue white yellow) In physics, vectors are things with a magnitude and a direction. Or, it is a point in a space. ( 4 ,3 ) vector addition, vector amplifications. We may depict a vector either as a row of numbers or a column of numbers. It is up to us, but we’ve to be consistent. u1 u 2 If u ( u1 ,u2 ,...,un ) then v u t is the .. u n transpose of the vector u . Conversely, v t u . Here, u row vector, v column vector Vector addition, scalar multiplication. Negation of a vector ( if v, then what is –v?). A vector u ( u1 ,u2 ,...,un ) is a tuple in an n-space. The components ui are also called coordinates, entries, elements, etc. One may consider a set of unit vectors comprising this R n ( or C n ) space. The corresponding vector- space that it collectively spans allows us to express each vector as follows: n u i1u1 i2 u 2 ... ik u k ... in u n ik u k k 1 If the vector space is orthogonal, each unit vector spanning the space is orthogonal to the other. The collection of the unit vectors that span completely any vector in the vector space is called a basis. In the above, the set of unit vectors { i j } form an orthonormal basis. e.g. A three-dimensional orthogonal space Norm (or Length) of a vector. The norm or length of a vector u R n is || u ||. n If u ( u1 ,u2 ,...,un ) then || u || ui2 i 1 u is a unit vector if || u || 1 . Therefore, v̂ v is a unit vector. || v || Accordingly, if u and v are two points in the nspace R n , the distance between them is PQ where n 2 PQ ( ui vi ) i 1 P u Q v if the notion of a distance is ‘meaningful’ in that space. In such a space, the dot product (or the inner product) between two vectors is defined in this way: For u ( u1 ,u2 ,u3 ,...,un ) and v ( v1 ,v2 ,...,vn ) , the dot product is u .v or u | v (this Dirac notation will be explained shortly) u .v u | v ( u1v1 u 2 v2 ... u n vn ) ui vi i In terms of the physics-metaphor u | v | u || v | cos , where is the angle between the two vectors. The norm of a vector u is then u | u . We should be careful. This is only one variety of norm that we can think of with a vector. In fact, we can have a number of them: n l1 norm : || u ||1 | ui | i 1 l 2 norm : || u ||2 || u || as we have defined it. n l p norm : || u ||p | ui | p i 1 l norm : || u || max | ui | i 1/ p For our sake, we’d mostly consider l2 norm . An inner product space V is a vector space where norm of a vector is defined and one could form a dot product u | v between any pair of vector u and v in it with the following conditions: 1. u | u 0 the length of a vector is never negative. 2. u | v v | u symmetry 3. u w | v u | v w | v 4. u | v u | v with a constant 5. u v | u v u | u v | v triangle inequality 6. u | v u | u v | v Cauchy-Schwarz inequality ex. u ( 2 0 - 1) v (1 3 - 2) u |u 4 0 1 5 v | v 1 9 4 14 and u | v (-2 1 0 3 (-1) (-2)) 0 5 14 ex. Let V R 2 be an inner product space where the dot products are defined in the following term: (( a ,b ),( c , d )) ac bd Then (( a ,b ),( a ,b )) a 2 b 2 0 (condition 1 is fulfilled and R 2 is an inner product space. ex. A polynomial space P n is an inner product space where every vector is an n-degree polynomial (polynomial-format?) like p( x ) a0 a1 x a2 x 2 ... an x n q( x ) b0 b1 x b2 x 2 ... bn x n Then p | q a0 b0 a1b1 ... anbn is an inner product between two such vectors in P n . Another inner product in P n may be defined as 1 p | q p( x )q( x )dx 0 ex. Another inner product space is the space of continuous functions where the dot product between f ( x ) and g ( x ) is defined as f |g f(t)g(t)dt - In this set up, the basis set S { v̂1 , v̂2 ,..., } is an infinite set of vectors like 1 1 1 cos t , v̂3 sin t , , v̂2 2 1 1 sin nt cos nt , v̂2 n 1 v̂1 v̂2 n That these form an orthonormal basis is evident from the fact that cos mt sin nt dt 0 for m n , cos mt cos nt dt 0 for m n sin mt sin nt dt 0 for m n ex. A weighted Eucledian inner product space R n may be designed to yield dot products like u | v 1u1v1 2u2 v2 ... nun vn with 1 2 ... n 1 The neural networks are developed on such spaces. More observation. A set S in an inner product space V is called orthogonal if any two distinct vectors in S are orthogonal. If also each vector is a unit vector, the set S is called orthonormal. The set of unit vectors in R n comprise an orthonormal space if i j | ik jk (Kronecker delta) jk 1 if j k, 0 otherwise Notice that a finite orthonormal set S is spanned by linearly independent vectors. Accordingly, if a set of vectors S V can be identified such that any vector v V can be expressed uniquely as a linear combination of vectors in S , the set S is a basis for V . Thus, v î1v1 î2 v2 î3 v3 ... în vn if each îk S . A vector space may have several distinct bases but each will have same number of basis vectors in them. The number of basis vectors spanning a vector space is called the dimension of a vector space. Ex. The standard basis spanning the R 3 is the set of three vectors S { e1 ,e2 ,e3 } where e1 ( 1 0 0) , e2 ( 0 1 0) and e3 ( 0 0 1) Any three-dimensional vector can be expressed in this basis. e.g. v ( 4 - 5 3) 4e1 5e2 3e3 Obviously other bases in this vector space are possible. Suppose, we choose these as our basis set S' { u1 ,u2 ,u3 } with u1 ( 1 - 2 1), u 2 (0 3 2) and u3 ( 2 1 - 1) Then, v ( 4 - 5 3) 1u1 2u2 3u3 and the coefficients are determined by the equations 1 2 3 4 2 1 3 2 3 5 1 2 2 3 3 Distances, angle and projections The distance between the two vectors (the distance between the terminals of the two vectors) d ( u ,v ) u | v If u and v are vectors in an inner product space V, the angle between these two vectors is given by cos u|v || u |||| v || The two vectors are orthogonal (perpendicular to each other) if u | v 0 Projection of a vector on another vector (see the diagram) u v u cos Projection of u on vector v is uv proj( u.v ) is <u.v> u.v uv proj( u.v ) u cos = ||u|| ||u||||v|| || v ||