Reading assignment Gilbert Strang Fast transforms: Banded matrices with banded inverses

D OMAGOJ M AJSTOROVIC Matrikelnummer: 03649878

November 17, 2013

Introduction
One of the most common digital-signal-processing techniques is the discrete Fourier transform (DFT), which breaks a signal into its component frequencies and can be represented as a matrix. Although the matrix for the Fourier transform is full, the great thing about the Fourier transform is that it happens to be possible, even though its full, to multiply fast and to invert it fast. However, for some signal-processing applications, banded matrices could prove more efcient than the Fourier transform. In this paper prof. Gilbert Strang describes a new way to split certain types of matrices into simpler matrices. The result could have implications for software that processes video, audio data or maybe for systems that control mechanical devices.

Banded Matrices
Prof. Strangs analysis applies to so-called banded matrices. Most of the numbers in a banded matrix are zeroes; the only exceptions fall along diagonal bands, at or near the central diagonal of the matrix. Since most of the entries in a banded matrix are zero, multiplying it by another matrix 1

is a very efcient procedure: You can ignore all the zero entries. However, the fact that a matrix is banded doesnt mean that its inverse is. B11 B12 0 B23 B33

B21 B22 0 B32 . .. . . . . ... . . 0

.. .. . . . B34 . .

B43 B44 B45 ... B54 B55 0 B65

0 B56 B66

0 . . . . . .

(1)

Figure 1: Tridiagonal 6-by-6 banded matrix with bandwidth 3 Actully the inverse of a banded matrix is almost always full, meaning that almost all of its entries are nonzero. Because of that speed advantages offered by banded matrices would be lost if restoring the signal required multiplying it by a full matrix. That is the reason why are the scientists interested in banded matrices with banded inverses - however, its not easy to know whether a banded matrix has banded inverse or not. Prof. Strang describes a new technique for breaking a banded matrix up into simpler matrices - matrices with fewer bands. Using this approach its easier to tell whether these simpler matrices have banded inverses, and if they do, their combination will, too.

Conclusion
This paper contribute to the analysis of one particular structure (perhaps the simplest): banded matrices. Prof. Strang showed very interesting tehnique which could be implemented in various cases of data sets which often yield matrices with regular structures (medical imaging, weather monitoring, sensing of all kinds, etc.). Bandedness is one type of structure, but there are others, and we can expect other mathematicians to apply techniques like Strangs to other types of structured matrices.