| Science Forum Index » Fractals Science Forum » fractal dimension for multidimensional data matlab program |
|
Page 1 of 1 |
|
| Author |
Message |
| sridhar reddy |
 Posted: Sat Nov 13, 2004 3:32 am |
|
|
|
Guest
|
hi everyone,
Iam trying to find the co-relational fractal dimension of
multidimensional cancer data set.The data is in the form of a matrix
144X16064 can anyone help me with a MATLAB program or any software
which would help me in finding fractal dimension for multidimensinoal
data.most of the programs work well for low dimensional data but fail
for my high dimensional data, waiting for your reply folks
sridhar |
|
|
| Back to top |
|
|
|
| sridhar reddy |
| Posted: Fri Nov 19, 2004 1:36 am |
|
|
|
Guest
|
sridhar.alluri@gmail.com (sridhar reddy) wrote in message news:<4a9625cb.0411130032.5e5bceb4@posting.google.com>...
[quote:0520a719e1]hi everyone,
Iam trying to find the co-relational fractal dimension of
multidimensional cancer data set.The data is in the form of a matrix
144X16064 can anyone help me with a MATLAB program or any software
which would help me in finding fractal dimension for multidimensinoal
data.most of the programs work well for low dimensional data but fail
for my high dimensional data, waiting for your reply folks
sridhar
[/quote:0520a719e1]
I know it is a little crazy to post a follow up message but i need
help regarding this so please anyone reply to my posting |
|
|
| Back to top |
|
|
|
| Tony Roberts |
| Posted: Tue Dec 07, 2004 9:08 pm |
|
|
|
Guest
|
sridhar reddy wrote:
[quote:2929423c6f]sridhar.alluri@gmail.com (sridhar reddy) wrote in message news:<4a9625cb.0411130032.5e5bceb4@posting.google.com>...
hi everyone,
Iam trying to find the co-relational fractal dimension of
multidimensional cancer data set.The data is in the form of a matrix
144X16064 can anyone help me with a MATLAB program or any software
which would help me in finding fractal dimension for multidimensinoal
data.most of the programs work well for low dimensional data but fail
for my high dimensional data, waiting for your reply folks
sridhar
I know it is a little crazy to post a follow up message but i need
help regarding this so please anyone reply to my posting
[/quote:2929423c6f]
Well first you have to explain more clearly what the data is, its
dimensionality and typical range and how obtained and how much data you
have.
Tony |
|
|
| Back to top |
|
|
|
| Guest |
| Posted: Wed Dec 08, 2004 6:06 pm |
|
|
|
|
hey toni
The data which iam currently using is microarray gene expression
data,it is basically a two dimensional matrix 16064 X 144, where the
144 columns are samples from different experimental conditions and each
sample has 16064 gene expression values,this data can be downloaded
from the link http://www.broad.mit.edu/cgi-bin/cancer/datasets.cgi
in that link iam using GCM_Training.res. I hope i answered all your
questions
sridhar |
|
|
| Back to top |
|
|
|
| Tony Roberts |
| Posted: Thu Dec 09, 2004 6:37 pm |
|
|
|
Guest
|
sridhar.alluri@gmail.com wrote:
[quote:48cc462ba7]hey toni
The data which iam currently using is microarray gene expression
data,it is basically a two dimensional matrix 16064 X 144, where the
144 columns are samples from different experimental conditions and each
sample has 16064 gene expression values,this data can be downloaded
from the link http://www.broad.mit.edu/cgi-bin/cancer/datasets.cgi
in that link iam using GCM_Training.res. I hope i answered all your
questions
sridhar
[/quote:48cc462ba7]
My guess is that you have no chance of detecting any fractal nature.
In essence you have 144 data points lying in a space of dimension
16,064. The dimension of the space is not necessarily the issue, but
see later. The first problem is that it is very hard to detect fractal
behaviour with only 144 points. My algorithms (see fdim at the
mathworks site unless it has disappeared) have a chance if the data is
clean and the dimensionality of the fractal is up to about 1.5. Other
algorithms will be hopeless.
But the real killer for you is that your microarray data will be very
noisy. Especially ruinous in the very high D space you are working
with: noise in each of 16,064 dimensions will fuzz completely any fine
scale fractal structure.
Give up the quest for fractals in this data,
Tony |
|
|
| Back to top |
|
|
|
| Guest |
| Posted: Fri Dec 10, 2004 1:05 am |
|
|
|
|
hey toni
the mit guys have given some cleaning techniques and filtering
techniques do you think applying them would be of any help, i will send
the exact link where they talk about filters.
http://www.broad.mit.edu/mpr/publications/projects/Global_Cancer_Map/PNAS_Supplementary_Information.pdf
After applying the filtering techniques and eliminating and setting
mean to 0 and vaience to one, the data set reduced to 144 samples with
10038 dimensions, and using the dime.m given by the MIT professor found
in google groups i got a fractal dimension more than 1.5, do you think
iam atleast heading in the right direction, and i have one more generic
doubt, given any other data set how do you know wether it is
self-smilar or not and how do we get an idea about the scales in which
they are self-similar.I hope iam not troubling you
sridhar |
|
|
| Back to top |
|
|
|
| Tony Roberts |
| Posted: Mon Dec 13, 2004 5:53 pm |
|
|
|
Guest
|
sridhar.alluri@gmail.com wrote:
[quote:473d7ca9eb]hey toni
the mit guys have given some cleaning techniques and filtering
techniques do you think applying them would be of any help, i will send
the exact link where they talk about filters.
http://www.broad.mit.edu/mpr/publications/projects/Global_Cancer_Map/PNAS_Supplementary_Information.pdf
After applying the filtering techniques and eliminating and setting
mean to 0 and vaience to one, the data set reduced to 144 samples with
10038 dimensions, and using the dime.m given by the MIT professor found
in google groups i got a fractal dimension more than 1.5, do you think
iam atleast heading in the right direction, and i have one more generic
doubt, given any other data set how do you know wether it is
self-smilar or not and how do we get an idea about the scales in which
they are self-similar.I hope iam not troubling you
sridhar
[/quote:473d7ca9eb]
No chance. You only have 144 data points. Any fractal dimension over
1ish has to be a reflection of the paucity of your data. If you got
some dimension less than 1, then you might have something. Give
fractals up for this data. Get productive using other methods.
Tony |
|
|
| Back to top |
|
|
|
| sridhar |
| Posted: Tue Dec 14, 2004 2:22 am |
|
|
|
Guest
|
hey toni,
I am very thankful for your suggestions.I just have a general doubt,I
hope this is a repeated question for you.Given any data set how do we
find out if it is self-similar or not.How do we find out the range in
which the data is self-similar.I would also be greatful if you could
send me a link of your fdim program
sridhar |
|
|
| Back to top |
|
|
|
| Tony Roberts |
| Posted: Wed Dec 15, 2004 6:32 pm |
|
|
|
Guest
|
sridhar wrote:
[quote:af6c63edb5]hey toni,
I am very thankful for your suggestions.I just have a general doubt,I
hope this is a repeated question for you.Given any data set how do we
find out if it is self-similar or not.How do we find out the range in
which the data is self-similar.I would also be greatful if you could
send me a link of your fdim program
sridhar
[/quote:af6c63edb5]
See http://www.sci.usq.edu.au/staff/robertsa/soft.html
A dataset is self-similar when you find a range in which some crucial
property shows self-similar behaviour. The range and the
self-similarity are two aspectes of the same question: you cannot have
one without the other.
Tony |
|
|
| Back to top |
|
|
|
|
