Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronic Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

LMS(Least Mean Square), anyone have infor on it?

Status
Not open for further replies.

ctsai23

Junior Member level 3
Joined
Jul 29, 2002
Messages
31
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
97
Does anybody have any information LMS (Least Mean Square)?
 

Humungus

Full Member level 6
Joined
Jul 10, 2001
Messages
384
Helped
41
Reputation
82
Reaction score
14
Trophy points
1,298
Activity points
3,985
It's a method to fit data. It can be linear or any function you want, provided that you only have two parameters to fit (offset and slope)
 

goodboy_pl

Full Member level 5
Joined
Mar 12, 2002
Messages
253
Helped
16
Reputation
32
Reaction score
15
Trophy points
1,298
Activity points
3,119
Also there is a method for controlling Adaptive filters with this name!
as Humungus said, it is a conventional method for linear regression and curve fitting, an important topic in statistic/math.
with this and other method of curve fitting it is possible to model systems from exprimental(measured) values of them. thus this method is widely used in device modeling(MOS,BJT,Caps,Inductors,...) in physics and other Advanced technologies.
I think the merit of this method in respect to other methods is simplicity and availability of analytical solution for it; also several functions which are not linear(such as log, exp) can be converted to linear function by some transforms.

BEST!
 

flatulent

Advanced Member level 5
Joined
Jul 19, 2002
Messages
4,629
Helped
489
Reputation
980
Reaction score
150
Trophy points
1,343
Location
Middle Earth
Activity points
46,689
more details

To expand on the previous good points, in LMS the difference between the data points and a continuous math function that is going to be used to approximate the data points are squared. This is the squared error. Then the sum of the squared errors at all of the data points is made. The parameters of the math function are varied until this sum of errors is made the smallest.

Even though it is easy to compute and find it has some problems. One is that large errors count more than small ones and positive ones count the same as negative ones. This may not be what you need in your approximation.
 

Status
Not open for further replies.
Toggle Sidebar

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Top