<< Chapter < Page Chapter >> Page >

So far, our approximation problem has been posed in an inner product space, and we have thus measured our approximation error using norms that are induced by an inner product such as the L 2 / 2 norms (or weighted L 2 / 2 norms). Sometimes this is a natural choice – it can be interpreted as the “energy” in the error and arises often in the case of signals corrupted by Gaussian noise. However, more often than not, it is used simply because it is easy to deal with.

In some cases we might be interested in approximating with respect to other norms – in particular we will consider approximation with respect to p -norms for p 0 . First, we introduce the concept of a “unit ball”. Any norm gives us rise to a unit ball, i.e., x : x = 1 . Some important examples of unit balls for the p norms in R 2 are depicted below.

Illustrations of the ell_p balls for p=1,2,infinity.  For p=2, the ball is a circle.  For p=infinity, it is a square.  For p=1, it is a diamond.

We now consider an example of approximating a point in R 2 with a point in a 1-D subspace while measuring error using the p norm for p = 1 , 2 , .

Suppose V = R 2 ,

A = span 2 - 1 , and x = 2 1 .

We will want to find x ^ A that minimizes x - x ^ p . Since x ^ A , we can write

x ^ = 2 α - α

and thus

e = x - x ^ = 2 - 2 α 1 + α .

While we can solve for α R to minimize e p directly in some cases, a geometric interpretation is also useful. In each case, on can imagine growing an p ball centered on x until the ball intersects with A . This will be the point x ^ A . that is closest to x in the p norm. We first illustrate this for the 2 norm below:

An illustration of approximation to a 1-D subspace in R2 using the ell_2 norm. The subspace is tangent to the sphere at the point where the radius of the sphere is just large enough to touch the line.  This provides another 'proof' of the orthogonality principle.

In order to calculate x ^ we can apply the orthogonality principle. Since e , [ 2 1 ] T = 0 we obtain a solution defined by α = 3 5 .

We now observe that in the case of the norm the picture changes somewhat. The closest point in is illustrated below:

An illustration of approximation to a 1-D subspace in R2 using the ell_infinity norm.  The square first touches the line at a different point than the sphere.  In this case both entries of the error vector have equal magnitude.

Note that the error is no longer orthogonal to the subspace A . In this case we can still calculate x ^ from the observation that the two terms in the error should be equal, which yields α = 1 3 .

The situation is even more different for the case of the 1 norm, which is illustrated below:

An illustration of approximation to a 1-D subspace in R2 using the ell_1 norm.  The diamond first touches the line at yet another point.  In this case the error vector has only one non-zero value.

We now observe that x ^ corresponds to α = 1 . Note that in this case the error term is [ 0 2 ] T . This punctuates a general trend: for large values of p , the p norm tends to spread error evenly across all terms, while for small values of p the error is more highly concentrated.

When is it useful to approximate in p or L p norms for p 0 ?

  • In some cases we will want the best fit to a specified frequency response in an L sense rather than the L 2 sense. This minimizes the maximum error rather than total energy in the error. In the figure below we illustrate a desired frequency response. If the L norm of the error is small, then we are guaranteed that the approximation to our desired frequency response will lie within the illustrated bounds.
    An illustration depicting a desired frequency response (low-pass) and maximum bounds by which we will allow other the actual frequency response to deviate from the desired one.
  • In compressing 3D geometry, can be useful to bound the L error to ensure that basic shapes of narrow features (like poles, power lines, etc.) are preserved.
  • In the case where the error is known to be sparse (i.e., zero on most indices) it can be useful to measure the error in the 1 norm.

Get Jobilize Job Search Mobile App in your pocket Now!

Get it on Google Play Download on the App Store Now




Source:  OpenStax, Digital signal processing. OpenStax CNX. Dec 16, 2011 Download for free at http://cnx.org/content/col11172/1.4
Google Play and the Google Play logo are trademarks of Google Inc.

Notification Switch

Would you like to follow the 'Digital signal processing' conversation and receive update notifications?

Ask