Is it possible to set specific OOBIndices in TreeBagger

I am trying to use TreeBagger for an image analysis algorithm in which the image is broken in n number of areas that I am trying to classify. I would like to specific OOBIndices so the OOB areas are from the same images. Meaning that the OOB randomization happens at the level of the image and not at the level of the area.

Thanks Walid

Answer 1

[Ilya]

To train TreeBagger, you pass X and Y. If "image" is your entire data X and if "areas" are your class labels Y, you already have that. OOB observations are drawn from the entire data, not from data for one specific class. If this is not what you mean, please clarify.

[Walid]

I will try to clarify: my X has 3 million observations coming from 1000 images. meaning on average 3000 observations per image.

right bow the treeBagger will proceed to randomly select observations (X)independently of the image dimension. I would like to set the OOB so they follow the image dimension and would dump large blocks of observations.

Hope this helps

[Ilya]

Not really. Do images represent classes in Y?

Answer 2

[Yiwen Mei]

I am facing a similar question. I want to set OOBIndices to specific elements rather than letting Matlab to randomly pick. Is it possible?

Thank you, Yiwen