Scribd
Upload
55 views2 pages

Adaptive Systems - Problem Class 5: The Leaky LMS Algorithm

This document discusses the leaky LMS adaptive filtering algorithm. It contains two problems: 1. The first problem analyzes the standard LMS algorithm when the input signal is a degenerated signal. It examines the autocorrelation matrix, eigenvalues, coefficient updates, and stability for different realizations of the input signal. 2. The second problem derives the leaky LMS algorithm from the leaky gradient descent algorithm. It analyzes the convergence point of the leaky gradient descent algorithm, derives a stability bound on the step size parameter, and calculates the convergence time constants.

Uploaded by

Khanh Nam
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
55 views2 pages

Adaptive Systems - Problem Class 5: The Leaky LMS Algorithm

This document discusses the leaky LMS adaptive filtering algorithm. It contains two problems: 1. The first problem analyzes the standard LMS algorithm when the input signal is a degenerated signal. It examines the autocorrelation matrix, eigenvalues, coefficient updates, and stability for different realizations of the input signal. 2. The second problem derives the leaky LMS algorithm from the leaky gradient descent algorithm. It analyzes the convergence point of the leaky gradient descent algorithm, derives a stability bound on the step size parameter, and calculates the convergence time constants.

Uploaded by

Khanh Nam
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

http://www.spsc.tugraz.

at/courses/adaptive

Adaptive Systems - Problem Class 5 The Leaky LMS Algorithm


Signal Processing and Speech Communication Laboratory Graz University of Technology Inffeldgasse 16c/II http://www.spsc.tugraz.at May 16, 2007

Problem 5.1 - Degenerated Input Signals and the LMS Algorithm


w[n] x[n] H (z ) [n] d

+
d[n]

c[n]

y [n]

e[n]

Figure 1: System identication. The input process is given by x[n] = 2cos(n + ), where is a random variable uniformly distributed between and . The unknown system and the adaptive FIR lter have the same order. The adaptive FIR lter has 2 coefcients. (a) Determine the auto-correlation matrix Rxx . (b) Determine the eigenvalues of Rxx . Can we calculate the Wiener solution copt ? (c) Consider one realization of the process x[n] for = 0. Use the LMS algorithm and visualize the coefcient update. Can we fully identify the unknown coefcient vector h? (d) Consider one realization of the process x[n] for = 0. Use the LMS algorithm and visualize the coefcient update. For a nite numerical range and precision, can the system become unstable?

http://www.spsc.tugraz.at/courses/adaptive

Problem 5.2 - The Leaky LMS Algorithm


We will investigate the leaky LMS algorithm given by c[n] = (1 )c[n 1] + e[n]x[n], where is the leakage parameter (0 < 1) and is the step-size parameter. Therefore, we will exploit the statistical relation to the leaky gradient descent algorithm: c[n] = (1 )c[n 1] + (p Rxx c[n 1]) (a) Starting from the leaky gradient descent algorithm, derive the leaky LMS algorithm by using instantaneous estimates for the autocorrelation matrix Rxx and the cross-correlation vector p. (b) Assume convergence, i.e., c[n] = c[n1] = c . Where does the leaky gradient descent algorithm converge to? (c) Introduce the misalignment vector v[n] and derive a stability bound on the step size parameter . (d) Derive the convergence time constants k .

You might also like