[Home] [Back Issues] [Search] [Vol 1] [Vol 2] [Vol 3] [Vol 4] [Vol 5]

The Wavelet Digest - Volume 1, Issue 3 (August 22, 1992)


Today's Editor: Wim Sweldens


Today's Topics:
   1. Papers available from Aware Inc.
   2. Papers available from University of Maryland
   3. Call for papers: Journal of Mathematical Imaging and Vision
   4. Announcement of a talk on wavelets
   5. Two wavelet problems
   6. Wavelet question
   7. Looking for a job in wavelets

Submissions for Wavelet Digest: 
    E-mail to wavelet@math.sc.edu with "submit" as subject.

Subscriptions for Wavelet Digest: 
    E-mail to wavelet@math.sc.edu with "subscribe" as subject.

Archive site:
    Anonyous ftp to maxwell.math.sc.edu (129.252.12.3),
    directory /pub/wavelet/archive.

Gopher server: bigcheese.math.sc.edu.

Current number of subscribers: 1804


From: P. Heller, Aware. Date: Mon, 10 Aug 92 10:42:33 -0400 Subject: Papers from Aware Inc. We at Aware have a full list of technical reports available for public distribution, including papers like Wayne Lawton, ``Necessary and sufficient conditions for constructing orthonormal wavelet bases" W. Lawton and H. Resnikoff, ``Multidimensional wavelet bases" David Pollen, ``Parametrization of compactly supported wavelets" P. Heller, H. Resnikoff, & R. O. Wells, ``Wavelet matrices and the representation of discrete functions" J. Weiss, W. Lawton, et. al. ``Wavelet analysis and the numerical solution of partial differential equations" and more. To request one of these papers or get a full list of the available papers, contact: Michael McPherson Aware, Inc. One Memorial Drive Cambridge, MA 02142 (617) 577-1700 michael@aware.com
From: P.S.Krishnaprasad, University of Maryland. Date: Thu, 20 Aug 92 16:02:54 EDT Subject: Papers available Following are abstracts of recent work related to wavelets and frames. Those interested in obtaining copies of the same may contact the undersigned. -------------------------------------------------------------------------- Professor P. S. Krishnaprasad | Tel: (301)-405-6843 Department of Electrical Engineering & | Fax: (301)-405-6707 Systems Research Center | A.V. Williams Building - Rm 2233 | Internet: University of Maryland | krishna@src.umd.edu College Park, MD 20742. | krishna@poincare.src.umd.edu -------------------------------------------------------------------------- SYSTEMS RESEARCH CENTER TECHNICAL REPORT, SRC-TR-90-44r1 University of Maryland, College Park, MD 20742. (also to appear in IEEE Transactions on Neural Networks) \centerline{\bf Analysis and Synthesis of Feedforward Neural Networks} \centerline{\bf Using Discrete Affine Wavelet Transformations} \centerline{by Y.C. Pati and P.S. Krishnaprasad} In this paper we develop a theoretical description of standard feedfoward neural networks in terms of discrete affine wavelet transforms. This description aids in establishing a rigorous understanding of the behavior of feedforward neural networks based upon the properties of wavelet transforms. Time-frequency localization properties of wavelet transforms are shown to be crucial to our formulation. In addition to providing a solid mathematical foundation for feedforward neural networks, this theory may prove useful in explaining some of the empirically obtained results in the field of neural networks. Among the more practical implications of this work are the following: (1) Simple analysis of training data provides complete topological definition for a feedforward neural network. (2) Faster and more efficient learning algorithms are obtained by reducing the dimension of the parameter space in which interconnection weights are searched for. This reduction of the weight space is obtained via the same analysis used to configure the network. Global convergence of the iterative training procedure discussed here is assured. Moreover, it is possible to arrive at a non-iterative training procedure which involves solving a system of linear equations. (3) Every feedforward neural network constructed using our wavelet formulation is equivalent to a 'standard feedforward network.' Hence properties of neural networks, which have prompted the study of VLSI implementation of such networks are retained. --------------------------------------------------------------- SYSTEMS RESEARCH CENTER TECHNICAL REPORT, SRC-TR-92-44, University of Maryland, College Park, MD 20742. (shorter version presented at Conference on Information Sciences and Systems, Princeton, March 1992) \centerline{\bf Affine Frames of Rational Wavelets in $H2 ( \Pi^+)$} \centerline{by Y.C. Pati and P.S. Krishnaprasad} In this paper we investigate frame decompositions of $H2(\Pi^+)$ as a method of constructing rational approximations to nonrational transfer functions in $H2(\Pi^+ )$. The frames of interest are generated from a single analyzing wavelet. We consider the case in which the analyzing wavelet is rational and show that by appropriate grouping of terms in a wavelet expansion, $H2(\Pi^+ )$ can be decomposed as an infinite sum of a rational transfer functions which are related to one another by dilation and translation. Criteria for selecting a finite number of terms from such an infinite expansion are developed using time-frequency localization properties of wavelets. ----------------------------------------------------------------- SYSTEMS RESEARCH CENTER Ph.D. Dissertation Series, Ph.D. 92-13 Title of Dissertation: Wavelets and Time-Frequency Methods in Linear Systems and Neural Networks Yagyensh C. Pati, Doctor of Philosophy, 1992 Dissertation directed by:Professor P. S. Krishnaprasad, Department of Electrical Engineering In the first part of this dissertation we consider the problem of rational approximation and identification of stable linear systems. Affine wavelet decompositions of the Hardy space H$^2(\Pi^+)$, are developed as a means of constructing rational approximations to nonrational transfer functions. The decompositions considered here are based on frames constructed from dilations and complex translations of a single rational function. It is shown that suitable truncations of such decompositions can lead to low order rational approximants for certain classes of time-frequency localized systems. It is also shown that suitably truncated rational wavelet series may be used as `linear-in-parameters' black box models for system identification. In the context of parametric models for system identification, time-frequency localization afforded by affine wavelets is used to incorporate {\em a priori} knowledge into the formal properties of the model. Comparisons are made with methods based on the classical Laguerre filters. The second part of this dissertation is concerned with developing a theoretical framework for feedforward neural networks which is suitable for both analysis and synthesis of such networks. Our approach to this problem is via affine wavelets and the theory of frames. Affine frames for L$^2$, are constructed using combinations of sigmoidal functions and the inherent translations and dilations of feedforward network architectures. Time-frequency localization is used in developing methods for the synthesis of feedforward networks to solve a given problem. These two seemingly disparate problems both lie within the realm of approximation theory, and our approach to both is via the theory of frames and affine wavelets.
From: Andrew Laine, University of Florida. Date: Thu, 20 Aug 92 15:47:42 -0400 Subject: Call for papers Dear Colleagues: An announcement for papers was made at this summer's NSF/SIAM wavelet conference for a special issue of the Journal of Mathematical Imaging and Vision, focusing on wavelet theory and applications. This message is to inform you that the call for papers has been extended beyond the initial deadline. Sincerely, Andrew Laine Guest Editor, Journal of Mathematical Imaging and Vision FINAL CALL FOR PAPERS: EXTENDED DEADLINE: September 15, 1992 Papers focusing on wavelets and their applications including but not limited to the topics listed below are solicited: - image pyramids - frames and overcomplete representations - multiresolution algorithms - wavelet-based noise reduction and restoration - multiscale edge detection - wavelet texture analysis and segmentation - wavelet-based fractal analysis - multiscale random processes - image representations from wavelet maxima or zero-crossings - wavelet coding and signal representation - wavelet theory and multirate filterbanks The aim of special issue is to emphasize the role of mathematics as a rigorous basis for imaging science. Emphasis will be on innovative or established mathematical techniques applied to vision and imaging problems in a novel way, including new developments (and problems) in mathematics arising from these applications. Expository articles, and short articles (no longer than 10 pages) are also encouraged. The special issue will be published in early 1993. Manuscripts should not exceed 30 pages double spaced. Please submit three copies of your manuscript to: Professor Andrew Laine Department of Computer and Information Sciences Computer Science and Engineering Building (CSE) Room 301 University of Florida Gainesville, FL 32611 Email: LAINE@CIS.UFL.EDU Phone: (904) 392-1239 Fax: (904) 392-1220
From: phillies@wpi.wpi.edu (George D. Phillies) Date: Thu, 20 Aug 1992 10:46:45 -0400 Subject: Talk. The following is the title and abstract for a lecture I will be giving at WPI on wavelet decompositions and their applications. The talk will be given on Moday, August 31, 1992, at 4PM in the Olin Hall of Physics. The treatment of wavelets will be strictly introductory, and aimed more at physicists than mathematicians. The new material will be the use of wavelet decompositions to study the Ising model of a magnet, and the sorts of information one can extract from Monte Carlo dynamics simulations via wavelet decompositions. We have just begun writing up our results, so preprints are not available. To reach WPI, take Route 290 to the Route 9-Highland Street exit, go west on Highland Street approximately 1 mile to West Street (traffic light), and three blocks north (right turn, if moving away from Route 290). Olin Hall is on the west (left) side of the street, the second building before West Street comes to an end. * * * * * WAVELET DECOMPOSITIONS: INTRODUCTION AND APPLICATION TO THE ISING MODEL George D. J. Phillies Department of Physics Worcester Polytechnic Institute First, I will introduce wavelet expansions, which are a new mathematical method that may be useful in physics. Second, I will show how a specific wavelet transform, the Burt-Adelson$^{1}$ decomposition, can be used to describe statics and dynamics of the Ising model. The Burt-Adelson decomposition was originally introduced as a tool for image processing; the Ising model is a highly-simplified statistico- mechanical description of a spin lattice, nominally similar to the lattices found in magnets. Wavelets may formally be viewed as extensions of the orthogonal functions familiarly used in physics. Isolated examples of wavelet expansions, such as Haar and Gabor transforms, have long been known. Recently, Mallat$^{2}$, Meyer$^{3}$, Daubechies$^{4}$ and others have shown that Haar and Gabor transforms are special cases of an extensive set of mathematical functions. Wavelets provide complete, sometimes non-orthogonal, sets of basis functions. Some wavelet expansions can be generated recursively, suggesting computation efficiencies. Other wavelet expansions have compact support, so that, e.g., in a wavelet decomposition of a piece of music, alteration of a single note changes local coefficients without affecting most of the expansion. While there is a large mathematical literature on wavelets, concrete applications of wavelet expansions in the physics literature do not appear to be common. I will describe Monte Carlo simulations (made by J. Stott and I last Summer) on static and dynamic properties of the 1-D Ising model, as revealed by wavelet expansions. 1) P. J. Burt and E. H. Adelson, IEEE Trans. on Communications 31, 532 (1983). 2) S. Mallat, {\em A Theory for Multiresolution Signal Decomposition: The Wavelet Transform } (see IEEE Trans.\ on Pattern Analysis and Machine Intelligence}, 11, 674-693 (1989). 3) Y. Meyer, {\em Ondelettes et Functions Splines}, Ecole Polytechnique (Paris) (1986). 4) Ingrid Debauchies, Commun.\ Pure and Applied Mathematics 41, 909-996 (1988).
From: Steve Lutes, Texas A&M University. Date: Wed, 19 Aug 92 21:21:30 CDT Subject: Two wavelet problems Hello! I've got two wavelet probelms for your consideration. They are: - First Problems-Detecting Abnormal EEG (brainwaves) Waveforms: Brain/CNS (Central Nervous System) pathologies often produce ab- normal waveforms in the EEG. Can wavelet techinques detect these wave- forms? Because EEG waveform interpetation is a specialized field with jargon, I've prepared a PostScript document explaining the jargon, giv- ing pictures of abnormal EEG waveforms, and explaining this problem in more detail. Please download and print this file before commenting. The file name is waveprob.ps. (Note from the editor: this file can be retrieved using anonymous ftp to maxwell.math.sc.edu, directory /pub/wavelet/announce.) - Second Problem: Making Wavelets from Data or Custom Wavelets. Can you make a wavelet from data? That is, can you take a waveform and change it into a wavelet? Addendum: I heard that you cam use splines to generate wavelets. Could you use splines to approximate a waveform, wavelet? Reply to N020BA@tamvm1.tamu.edu.
From: Ming-Haw Yaou, National Chiao Tung University Taiwan. Date: Mon, 17 Aug 92 16:18:48 CST Subject: Wavelet question. In the wavelet digest vol.1 nr.1 (Topic #6), I asked a question about orthonormal wavelet transform. It is very kind of many readers to email me their constructive advice. I also have to apologize for that some misunderstanding was happened due to my unclear expression in the problem. To specify my question more clearly, I correct some mistakes in the properties (a) and (b) and express them in explicit equation. ------------------ The rewritten properties (a) and (b) ------------------ (a) Both Pmn(x) and Qmn(x) are orthonormal bases. This can be written as < P(x-n) , P(x-m) > = delta(m,n). < Q(x-n) , Q(x-m) > = delta(m,n). Pmn(x) and Qmn(x) are mutually orthogonal (under the same value of m). This can be expressed as < P(x-n) , Q(x-m) > = 0. (b) Denote the second derivatives of P(x) and Q(x) as P"(x) and Q"(x). P"(x) and Q"(x) satisfy < P"(x-n) , Q"(x-m) > = 0. (c) Further constraints (for avoiding trivial solution) : The regions of support of P(x) and Q(x) should not be non-overlapped. Also the regions of support of P(x) and Q(x) should be greater than 1, i.e., in the interval [p,q] and |p-q| > 1. Where operation < a , b > stands for the inner product of continuous functions a with b. Symbol delta(m,n) stands for the Kronecker delta function. Variables m and n are integers and x are real. The constraints in property(a) are well known as the necessary conditions in the construction of dyadic orthonormal wavelet bases. The property(b) came from my current study ( I wish to solve curve fitting problem from wavelet theory). PS: It is known that the property(a) can be indirectly solved by transferring the constraints to the filters of an associated QMF banks. And P(x) and Q(x) can be iteratively calculated from the filters. I try to transfer the additional property(b) to the filters. But It seems not to be very easy. I will very appreciate if any answer or advice is sent from you. Please e-mail to : u7811837@cc.nctu.edu.tw Thank in advance, Ming-Haw Yaou.
From: Evan Greenberg, University of Massachusetts. Subject: Looking for a job in wavelets. I recently completed my M.S. in Applied Mathematics at the University of Massachusetts, Amherst. I'm currently seeking employment in the mathematical sciences, preferably along the lines of research or applications/research. My undergraduate education is a B.S.E.E. from Tufts University. As Wavelets seems to be an area which holds great promise for the future (and present) in both pure and applied areas, I'd be very enthusiastic about a good opportunity for wavelet-related work where I could also gain more knowledge on some specific topics. I currently have some basic background in wavelets and multiresolution analysis; my interests also include Kalman filter theory and signal processing. Ideally, I'd like to be able to work in an environment that provides both flexibility and good support for learning. Nothing would be please me more than being able to actually work with mathematics that I've learned in grad school. Although I'd love to get a Ph.D., I'd rather be employed full-time now. Wavelets are in a diverse variety of applications now, but I'm especially interested in their use with audio compression technology and state-of-the-art recording. I have some work experience in non-mathematical engineering, during the time between my bachelor's and master's degrees. I'd welcome any insight into how I might find worthwhile employment, in addition to any actual job opportunites. email: greenber@math.umass.edu phone (413) 253-0555 address: evan greenberg, 216 bel-air drive, longmeadow, ma 01106

<webmaster@wavelet.org>