3d graph showing reconstruction error

Question :

I want to plot 3 axis giving me 3d graph ! One axis is error, second is number of iteration, 3rd is % of missing sample

My Conjugate gradient based algorithm is giving me fully reconstructed sample i put in the signal missing samples.

So assuming this is my example

A= signal with missing sample
P=% of missing samples in A
b=reconstructed signal
n=number of iteration
e=error

so i put in signal with 512 samples with 5% missing samples and it takes 20 iteration with say 0.002 as error...

now i will increase the missing part to 10% and it will obviously take more iteration and error may be increase depending on missing samples.

I want to do it from 5% till 70% missing samples...

Looking for your ideas for input

Cheerr

kk

Answer 1

[Andrew Newell]

If I understand your question, you have two input variables, P and n, and a result e = e(P,n). So you create a grid of values for P and n using meshgrid, then calculate e for each pair P(i),n(j). If the resulting error is a smooth function of the inputs, you could plot it using

surf(P,n,e)

or use contour. If it's not so smooth, you might prefer to use imagesc.

EDIT: To clarify, you need a loop something like this:

p = 5:5:70;
n = 20:20:200;
[P,N] = meshgrid(p,n);
e = 0*P; % just an initialization
for i=1:numel(P)
   e(i) = simulateAndReturnError(P(i),N(i));
end
surf(P,N,e) % or imagesc(p,n,e)

[guj]

Here it goes ! After running simulation i get error results as 1 by 3 vector , when i change parameter i get 1 by 5 vectors they keep on increasing maximum till 1000 values with change of parameter (which i will change 13 times) so my question is

Suppose y is error i am getting after every simulation with number of values less than 1000, so i want to preallocate the vector say y1...which will store value of y after every simulation...so there are 13 simulation...I want to make a matrix of 13 by 1000;

So I have solution which is

error matrix (y1)of 13 by 1000

vector of 13 parameter values....1 by 13

iteration matrix : 1 by 1000 so basically error matrix is what i get after changing parameter 13 times and running 1000 iteration each...

Question 2:

Now I want to plot all three parameters on 3d axis ! so should i do like this

[X,Y]=meshgrid(P,I) where p is parameter vector 1 by 13 and I is 1 by 1000.

surf(X,Y,errormatrix)

Thanks

[Andrew Newell]

What is question 1?

[Andrew Newell]

What you plot depends on how you index errormatrix (which is why I provided the code). You should make sure that errormatrix(i,j) corresponds to P(j) and I(i), e.g., errormatrix(100,2) is for 100 iterations and your second parameter. If it's the other way around, use:

surf(X,Y,errormatrix')

[guj]

THanks I am able to plot it but stuck with some thing else now ..posted another question in link on using surf and meshgrid