[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## 2D cluster of 2D matrix in Octave?

**From**: |
Mark Wall |

**Subject**: |
2D cluster of 2D matrix in Octave? |

**Date**: |
Thu, 18 Apr 2002 18:48:22 -0500 |

Hello!

`I would love to use Octave for the following 2 tasks, but am unable
``to determine if Octave will work. My primary question- is this
``possible in Octave? Secondary- if yes, how?
`

`We routinely use MatLab with the Statistics Tool Box to manipulate
``and plot a 2 dimensional matrix (m x n) of real numbers. If you
``think of each dimension as an array of vectors (e.g. a collection of
``m vectors (each n-dimensions), we 1) cluster these (and generate a
``dendrogram) for both dimensions as below using a city-block distance
``measure:
`
original matrix with labels
----------------
W X Y Z
A 0 1 0 0
B 0 1 0 1
C 5 0 0 0
D 0 0 0 2
E 5 0 0 0
Clustered matrix
----------------
W Y X Z
E 5 0 0 0
C 5 0 0 0
A 0 0 1 0
B 0 0 1 1
D 0 0 0 2

`In reality, m and n are between 50 and 1000 so we 2) display these
``matrices as an x-y-color plot for easy visualization. Think of this
``as a square frame with each pixel representing a matrix element. The
``values are represented by a color continuum (dark blue = low values,
``through yellow to red = high values). I can provide a JPEG if
``necessary.
`

`With large matrices, we isolate sub-clusters by creating new matrices
``that are the appropriate slices of the original matrix and then
``repeat steps 1 and 2 on the sub-matrices.
`
Thank you,
Mark Wall
---------
HHMI/UT Southwestern Medical Center
5323 Harry Hines Blvd.
Dallas, Texas 75390-9050
USA
214.648.5050
-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.
Octave's home on the web: http://www.octave.org
How to fund new projects: http://www.octave.org/funding.html
Subscription information: http://www.octave.org/archive.html
-------------------------------------------------------------

**2D cluster of 2D matrix in Octave?**,
*Mark Wall* **<=**