Normed and Inner Product Spaces Banach and Hilbert Spaces

Normed and Inner Product Spaces Banach and Hilbert Spaces

Mathematics for Control Theory Elements of Analysis p.2 Normed and Inner Product Spaces Banach and Hilbert Spaces Hanz Richter Mechanical Engineering Department Cleveland State University Reading materials We will use: ■ Anthony N. Michel and Charles J. Herget (2007) [1981], Algebra and Analysis for Engineers and Scientists, Birkhäuser, e-ISBN-13: 978-0-8176-4707-0 (chapter 6, sections 6.1 through 6.5, 6.11 and 6.15A) 2 / 21 Normed Linear Spaces Let X be a Linear vector space over some field F. A mapping from X into R is called a norm on X (denoted ||x|| for all elements x ∈ X) if ∀x, y ∈ X, ∀α ∈ F: 1. ||x|| ≥ 0 2. ||x|| =0 ⇐⇒ x =0 3. ||αx|| = |α|||x|| 4. ||x + y|| ≤ ||x|| + ||y|| The pair {X; ||.||} is called a normed linear space. Note that 4. implies n n || xi|| ≤ ||xi|| Xi=1 Xi=1 3 / 21 Normed Linear Spaces Induce Metric Spaces Let {X; ||.||} be a normed linear space. Define a function ρ : X × X 7→ R by ρ(x, y) = ||x − y|| for all x, y ∈ X. Then Theorem 6.1.2 in Michel and Herget indicates that {X; ρ} is a metric space. Prove this as an exercise. Observe the following: 1. We can interpret a normed linear space as a metric space with metric ρ(x, y) = ||x − y||. 2. Therefore, all properties that hold for metric spaces also hold for normed linear spaces. 3. Only an adjustment to the terminology is necessary. For example, an open ball with center x0 and radius r in a normed linear space is defined by: S(x0, r) = {x ∈ X : ||x − x0|| < r} 4 / 21 Banach Spaces A complete normed linear space is called a Banach space. That is, {X; ||.||} is defined to be a complete normed linear space iff {X; ρ} is a complete metric space, with ρ(x, y) = ||x − y||. Examples: n 1. Consider X = C . For x =(ζ1,ζ2,...ζn) ∈ X define the p-norm as n 1/p p ||x||p = |ζi| ! Xi=1 with 1 ≤ p < ∞. The corresponding p-metric was studied earlier, and we saw that it defined n a complete metric space. Therefore {C ; ||.||p} is complete. 5 / 21 Examples... 2. Define ||x||∞ = max |ζi| i The corresponding metric was also studied earlier and it defines a n complete metric space. Therefore {C ; ||.||∞} is complete. 3. lp spaces are frequently found in the analysis of discrete-time control systems. It’s important to use a lower-case l. Denote the set of infinitely-long vectors (sequences) of complex (or real in particular) ∞ numbers as X = C . The set lp is a linear subspace of X defined by: n p lp = {x ∈ X : |ζi| < ∞} Xi=1 The lp norm of any sequence of lp is defined by n 1/p p ||x||p = |ζi| ! Xi=1 6 / 21 Examples... 4. {lp; ||.||p} are also Banach spaces for 1 ≤ p < ∞. 5. Also define l∞ = {x ∈ X : sup |ζi| < ∞} i and ||x||∞ = sup {|ζi|} i Again, {l∞; ||.||∞} is a Banach space. 7 / 21 Completeness/Incompleteness of certain function spaces Consider X = C[a, b]. We know X is a linear vector space. Define the norms b 1/p p ||x||p = |x(t)| dt Za for all x ∈ X and 1 ≤ p < ∞. Although it can be verified that {C[a, b]; ||.||p} are normed linear spaces, we saw that the corresponding metric spaces are not complete, so these normed spaces are not Banach spaces. However, taking ||x||∞ = sup |x(t)| t∈[a,b] does correspond to a complete metric space. Hence, {C[a, b]; ||.||∞} is a Banach space. 8 / 21 Product Normed Spaces Paralleling the discussion of product metric spaces, suppose {X; ||.||x} and {Y ; ||.||y} are normed spaces. How do we define a norm for X × Y ? First, X × Y must be a linear space. This can be accomplished with the following addition and multiplication rules: (x1, y1)+(x2, y2)=(x1 + x2, y1 + y2) α(x, y)=(αx,αy) Then a valid norm for X × Y is simply ||(x, y)|| = ||x||x + ||y||y If the individual spaces are Banach, so will be their product under this norm. 9 / 21 Some Convexity Results Let X be a real normed linear space. Given x, y ∈ X, define the segment joining x and y by the set: xy = {z ∈ X : z = αx + (1 − α)y 0 ≤ α ≤ 1} Let Y ⊂ X. We say that Y is convex if xy ⊂ Y ∀x, y ∈ Y . 1. Let S be any family of convex sets. Then Y ∈S is also convex. 2. Definition: The convex hull of Y is the intersectionT of all convex sets which contain Y . It is the smallest convex set containing Y . 3. The closure of Y is convex whenever Y is convex. 10 / 21 Continuity of norm and bounded functionals The norm is a function from X into R. Theorem 6.1.15 shows that this function is continuous on X. Recall that a linear functional f is a mapping from a linear space to the reals: f : X 7→ R A linear functional must satisfy superposition and homogeneity, as seen in the previous handout. A functional f is bounded if ∃M ≥ 0 s.t. |f(x)| ≤ M||x|| for all x ∈ X. For linear functionals, continuity and boundedness are the same thing, as seen in Th. 6.5.5. Next, we define a norm for bounded linear functionals. 11 / 21 A norm for bounded functionals Let Xf be the linear space of all linear functionals on X and let X∗ be the subset of Xf formed bounded functionals. Define a mapping from X∗ into R by: |f(x)| ||f|| = sup x6=0 ||x|| for all f ∈ X∗. Then (Th. 6.5.6): 1. X∗ is a linear subspace of Xf 2. ||.|| is a norm on X∗ 3. {X∗; ||.||} is a Banach space Note that X∗ is Banach even if X isn’t. Space X∗ is called the normed dual of X, or just the dual of X. 12 / 21 Example Take X = Rn. Let a ∈ X be fixed and define the functional f : X 7→ R by the rule f(x) = aT x for all x ∈ Rn. 1. Show that f is a bounded linear functional 2. Find a formula for ||f|| using the of the previous slide. 13 / 21 Alternative functional norm formulas See Th. 6.5.10: Let f be a bounded linear functional and ||f|| its norm. Then ||f|| = inf {M : |f(x)| ≤ M||x|| ∀x ∈ X} M ||f|| = sup {|f(x)|} ||x||≤1 ||f|| = sup {|f(x)|} ||x||=1 Examine the norms for functionals on C[a, b] in Examples 6.5.12-13 carefully. Skip to Section 6.11. 14 / 21 Signal Norms of Interest in Control In control systems analysis, it is important to establish the “size” of various signals. For instance, if we restrict the class of signals to continuous functions defined on finite intervals, we saw that b 1/p p ||x||p = |x(t)| dt Za is a valid norm. A few limitations so far: 1. An infinite value of the norm could arise if we extend b to infinity. But there is a subclass of functions for which the integral is still finite. 2. The normed space {C[a, b]; ||.||p} is not complete for 1 ≤ p < ∞. Cauchy sequences where every element is a continuous function could fail to converge to a continuous function. 3. Signals found in engineering systems must include discontinuous functions. These limitations motivate extension to a wider class of functions. The set of Lebesgue measurable functions will be outlined later in the course. 15 / 21 Inner Product Spaces Let X be a linear space over C. An inner product on X is a function h.i : X × X 7→ C satisfying: 1. hx,xi > 0 for all x 6=0 and hx,xi =0 if x =0 (positive-definite function). 2. hx, yi = hy,xi for all x, y ∈ X 3. hαx + βy,zi = αhx,zi + βhy,zi for all x,y,z ∈ X and all α,β ∈ C (bilinearity) 4. hx,αy + βzi =α ¯hx, yi + β¯hx,zi for all x,y,z ∈ X and all α,β ∈ C The combination {X; h.,.i} is called an inner product space. 16 / 21 Inner Product Spaces Induce Normed Spaces Notational simplification: refer to the inner product space {X; h.,.i} in context as just X. Let X be an inner product space. Then the function ||.|| : X 7→ R defined by ||x|| = hx,xi1/2 is a norm (satisfies the 4 conditions presented earlier). Also, the Cauchy-Schwarz inequality holds: |hx, yi| ≤ ||x||||y|| We can associate a normed linear space to every inner product space by using the above norm. As done before, we can in turn associate a metric space. All properties that hold for a metric spaces will hold for inner product spaces. A complete inner product space is called a Hilbert space. Inner product spaces including incomplete ones are called pre-Hilbert spaces. 17 / 21 Important Hilbert Spaces ■ Cn with any suitable inner product is a Hilbert space. Note: Let x, y ∈ C be represented as column vectors with n components. The proper way to compute their inner product is x¯T y and not just xT y. In Matlab: x’*y will conjugate as well as transpose x . If for whatever reason you want to block the conjugation, use x.’*y ■ The space of summable sequences lp studied earlier: Let x =(ζ1,ζ2,..) and y =(η1, η2,...) be 2 sequences in lp.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    21 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us