11/19/2011
Lecture 7
Least Mean Square (LMS)
Adaptive Filtering
Dr. Tahir Zaidi
Steepest Descent
The update rule for SD is
where
or
SD is a deterministic algorithm, in the sense that p and R are
assumed to be exactly known.
In practice we can only estimate these functions.
Chapter 5
Adaptive Signal Processing
11/19/2011
Basic Idea
The simplest estimate of the expectations is
To remove the expectation terms and replace them with the
instantaneous values, i.e.
Then, the gradient becomes
Eventually, the new update rule is
Chapter 5
No
expectations,
Instantaneous
samples!
Adaptive Signal Processing
Basic Idea
However the term in the brackets is the error, i.e.
then
is the gradient of
instead of
as
in SD.
Chapter 5
Adaptive Signal Processing
11/19/2011
Basic Idea
Filter weights are updated using instantaneous values
Chapter 5
ELE 774 - Adaptive Signal Processing
Update Equation for
Method of Steepest Descent
Update Equation for
Least Mean-Square
Chapter 5
ELE 774 - Adaptive Signal Processing
11/19/2011
LMS Algorithm
unbiased
Since the expectations are omitted, the estimates will have a high variance.
Therefore, the recursive computation of each tap weight in the LMS
algorithm suffers from a gradient noise.
In contrast to SD which is a deterministic algorithm, LMS is a member of the
family of stochastic gradient descent algorithms.
LMS has higher MSE (J()) compared to SD (Jmin) (Wiener Soln.) as n
i.e., J(n) J() as n
Difference is called the excess mean-square error Jex()
The ratio Jex()/ Jmin is called the misadjustment.
Hopefully, J() is a finite value, then LMS is said to be stable in the
mean square sense.
LMS will perform a random motion around the Wiener solution.
Chapter 5
Adaptive Signal Processing
LMS Algorithm
Involves a feedback connection.
Although LMS might seem very difficult to work due the
randomness, the feedback acts as a low-pass filter or performs
averaging so that the randomness can be filtered-out.
The time-constant of averaging is inversely proportional to .
Actually, if is chosen small enough, the adaptive process is made
to progress slowly and the effects of the gradient noise on the tap
weights are largely filtered-out.
Computational complexity of LMS is very low very attractive
Only 2M+1 complex multiplications and 2M complex additions
per iteration.
Chapter 5
Adaptive Signal Processing
11/19/2011
LMS Algorithm
Chapter 5
Adaptive Signal Processing
Canonical Model
LMS algorithm for complex signals/with complex coef.s can be
represented in terms of four separate LMS algorithms for real
signals with cross-coupling between them.
Write the input/desired signal/tap gains/output/error in the complex
notation
Chapter 5
Adaptive Signal Processing
10
11/19/2011
Canonical Model
Then the relations bw. these expressions are
Chapter 5
Adaptive Signal Processing
11
Canonical Model
Chapter 5
Adaptive Signal Processing
12
11/19/2011
Canonical Model
13
Adaptive Signal Processing
Chapter 5
Analysis of the LMS Algorithm
Although the filter is a linear combiner, the algorithm is highly nonlinear and violates superposition and homogenity
Assume the initial condition
Analysis will continue using the weight-error vector
, then
output
and its autocorrelation
Chapter 5
input
Here we use expectation,
however, actually it is
the ensemble average!.
Adaptive Signal Processing
14
11/19/2011
Analysis of the LMS Algorithm
We have
Let
Then the update eqn. can be written as
Analyse convergence in an average sense
Algorithm run many timesstudy their ensemble average behavior
Chapter 5
Adaptive Signal Processing
15
Small Step Size Analysis
Assumption I: step size is small (how small?) LMS filter act
like a low-pass filter with very low cut-off frequency.
Assumption II: Desired response is described by a linear multiple
regression model that is matched exactly by the optimum Wiener
filter
where eo(n) is the irreducible estimation error and
Assumption III: The input and the desired response are jointly
Gaussian.
Chapter 5
Adaptive Signal Processing
16
11/19/2011
Small Step Size Analysis
Applying the similarity transformation resulting from the eigendecom.
on
We do not have this term in
Wiener filtering!.
i.e.
Then, we have
where
Components of v(n)
are uncorrelated!
17
Adaptive Signal Processing
Chapter 5
Small Step Size Analysis
Components of v(n) are uncorrelated:
stochastic force
first order difference equation (Brownian motion, thermodynamics)
Solution: Iterating from n=0
natural component
of v(n)
Chapter 5
forced component
of v(n)
Adaptive Signal Processing
18
11/19/2011
Learning Curves
Two kinds of learning curves
Mean-square error (MSE) learning curve
Mean-square deviation (MSD) learning curve
Ensemble averaging results of many () realizations are averaged.
What is the relation bw. MSE and MSD?
for small
Chapter 5
19
Adaptive Signal Processing
Learning Curves
for small
under the assumptions of slide 17.
Excess MSE
LMS performs worse than SD, there is always an excess MSE
use
Chapter 5
Adaptive Signal Processing
20
10
11/19/2011
Learning Curves
or
Mean-square deviation D is lower-upper bounded by the excess MSE.
They have similar response: decaying as n grows
Adaptive Signal Processing
Chapter 5
21
Convergence
For small
Hence, for convergence
or
The ensemble-average learning curve of an LMS filter does not
exhibit oscillations, rather, it decays exponentially to the const. value
Jex(n)
Chapter 5
Adaptive Signal Processing
22
11
11/19/2011
Misadjustment
Misadjustment, define
For small , from prev. slide
or equivalently
but
then
Adaptive Signal Processing
Chapter 5
23
Average Time Constant
From SD we know that
but
then
Chapter 5
Adaptive Signal Processing
24
12
11/19/2011
Observations
Misadjustment is
directly proportional to the filter length M, for a fixed mse,av
inversely proportional to the time constant mse,av
Directly proportional to the step size
slower convergence results in lower misadjustment.
smaller step size results in lower misadjustment.
Time constant is
inversely proportional to the step size
smaller step size results in slower convergence
Large requires the inclusion of k(n) (k1) into the analysis
Difficult to analyse, small step analysis is no longer valid,
learning curve becomes more noisy
Adaptive Signal Processing
Chapter 5
25
LMS vs. SD
Main goal is to minimise the Mean Square Error (MSE)
Optimum solution found by Wiener-Hopf equations.
Requires auto/cross-correlations.
Achieves the minimum value of MSE, Jmin.
LMS and SD are iterative algorithms designed to find wo.
SD has direct access to auto/cross-correlations (exact measurements)
can approach the Wiener solution wo, can go down to Jmin.
LMS uses instantenous estimates instead (noisy measurements)
Chapter 5
fluctuates around wo in a Brownian-motion manner, at most J().
Adaptive Signal Processing
26
13
11/19/2011
LMS vs. SD
Learning curves
SD has a well-defined curve composed of decaying exponentials
Chapter 5
For LMS, curve is composed of noisy- decaying exponentials
27
Adaptive Signal Processing
LMS Limits
As filter length increases, M, Imposes a limit on the step size to
avoid instability
If the upper bound is exceeded, instability is observed.
Smax: maximum component
of the PSD S() of the tap
inputs u(n).
Chapter 5
Adaptive Signal Processing
28
14
11/19/2011
LMS Example
One tap predictor of order one AR process. Let
Chapter 5
Adaptive Signal Processing
29
LMS Example
Chapter 5
Adaptive Signal Processing
30
15
11/19/2011
LMS Example
Chapter 5
Adaptive Signal Processing
31
LMS Example: Learning Curves
Chapter 5
Adaptive Signal Processing
32
16
11/19/2011
H Optimality of LMS
A single realisation of LMS is not optimum in the MSE sense
Ensemble average is.
The previous derivation is heuristic
(replacing auto/cross correlations with their instantenous estimates.)
In what sense is LMS optimum?
It can be shown that LMS minimises
Minimising the maximum of something minimax
Chapter 5
Maximum energy gain of the filter under the constraint
Optimisation of an H criterion.
Adaptive Signal Processing
33
H Optimality of LMS
Provided that the step size parameter satisfies the limits on the
prev. slide, then
no matter how different the initial weight vector
is from the
unknown parameter vector wo of the multiple regression model, and
irrespective of the value of the additive disturbance n(n),
the error energy produced at the output of the LMS filter will never
exceed a certain level.
Chapter 5
Adaptive Signal Processing
34
17
11/19/2011
Limits on the Step Size
Chapter 5
Adaptive Signal Processing
35
18
